{"articles":[{"author":"FirewallSync Editorial","description":"How session hijacking actually works after MFA succeeds — infostealer malware, XSS-based cookie theft, and AiTM relay — and why each bypasses MFA differently.","faq":[],"markdown_url":"https://firewallsync.com/posts/how-session-hijacking-works-even-after-mfa/index.md","published":"2026-11-21","reading_minutes":4,"summary":"MFA protects the login. It says nothing about the session token issued afterward — and several distinct attack techniques exist specifically to steal that token instead of touching the login at all.","tags":["authentication","incident response"],"takeaways":[],"title":"How Session Hijacking Works Even After MFA","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-11-21","url":"https://firewallsync.com/posts/how-session-hijacking-works-even-after-mfa/"},{"author":"John Marshall","description":"The European Commission confirmed an AI-generated cyber campaign, detected October 22, took several apps offline. No staff data breach found so far.","faq":[{"a":"The Commission says a cyber campaign that misused a publicly available AI tool was detected on October 22, 2026 and confirmed on October 26. As a protective measure it took several applications offline, and most have since been restored. The investigation is ongoing with CERT-EU and the European Data Protection Supervisor.","q":"What happened in the European Commission cyberattack?"},{"a":"The Commission says that at this stage of the investigation it has not detected any evidence of a data breach affecting its staff as a whole. Politico reported that the attackers potentially compromised large amounts of data. That report has not been confirmed by the Commission.","q":"Was data stolen from the European Commission?"},{"a":"The Commission has not named it. Spokesperson Thomas Regnier said only that it was a publicly available AI tool. Politico reported that CERT-EU detected the attackers using ARTEX, an AI penetration-testing framework, but the Commission has not confirmed this.","q":"Which AI tool was used against the European Commission?"},{"a":"No one has been named. Regnier said he would not go into attribution and did not answer a reporter's question about whether the tool was Chinese. As of November 21, 2026, no attribution appears in the Commission's on-record statements.","q":"Who is behind the attack on the European Commission?"}],"markdown_url":"https://firewallsync.com/posts/european-commission-ai-cyberattack-what-we-know/index.md","published":"2026-11-21","reading_minutes":6,"summary":"The European Commission says an 'AI generated cyber campaign' using a publicly available AI tool forced it to take several applications offline after detection on October 22. Here is what the Commission confirmed, what is only reported, and what engineers can check.","tags":["hacking news","european commission","ai security","artex","cert-eu","incident response"],"takeaways":["Commission spokesperson Thomas Regnier said the incident was detected on October 22 and confirmed on October 26, and called it an 'AI generated cyber campaign' that misused a publicly available AI tool.","The Commission temporarily took several applications offline as a protective measure and says most have been restored. It has not said which services were hit.","The Commission says it has found no evidence of a data breach affecting its staff 'at this stage of the investigation'. It declined to attribute the attack and said it was not aware of a ransom demand.","Politico reported that CERT-EU saw the attackers using ARTEX, an AI penetration-testing framework, and that affected services included the HR platform Sysper. The Commission has not confirmed either point.","Published guidance on ARTEX activity from SentinelOne includes alerting on user agents containing 'artex', keeping admin interfaces off the internet and rotating exposed secrets."],"title":"European Commission AI Cyberattack: What It Has Confirmed","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-21","url":"https://firewallsync.com/posts/european-commission-ai-cyberattack-what-we-know/"},{"author":"Craig Leslie","description":"CSAF VEX remediations and threats: 5 remediation categories, restart_required, impact threats per product, and why Trivy v0.76.2 rejects product_groups.","faq":[{"a":"CSAF 2.0 defines five: vendor_fix (an official fix from the product's author), workaround (a configuration or deployment change that avoids exposure), mitigation (reduces risk without resolving the flaw), none_available (no fix or remediation yet) and no_fix_planned (no fix will ever be provided). The CSAF 2.1 Committee Specification adds fix_planned and optional_patch.","q":"What are the CSAF remediation categories?"},{"a":"For each product in known_not_affected, the VEX profile requires either a machine-readable flag in vulnerabilities[].flags or a threat with category impact whose details say why the vulnerability cannot be exploited. Check that the flag or threat lists that product, directly or through a product group.","q":"Where does a CSAF VEX file explain why a product is not affected?"},{"a":"No. In Trivy v0.76.2 the CSAF matcher acts only on known_not_affected and fixed in product_status. It reads threats only to copy the first impact threat's details into the modified finding. Remediations, including vendor_fix URLs and restart_required, are never read.","q":"Does Trivy use CSAF remediations?"},{"a":"Three causes we reproduced against gocsaf v3.7.0, the library Trivy v0.76.2 uses: the file uses product_tree.product_groups, which gocsaf expects as an object rather than an array; the file declares csaf_version 2.1; or a remediation uses a CSAF 2.1 category such as fix_planned. Each makes JSON decoding fail.","q":"Why does Trivy fail to load my CSAF VEX file?"}],"markdown_url":"https://firewallsync.com/posts/csaf-vex-remediations-threats-category-product-ids-trivy/index.md","published":"2026-11-21","reading_minutes":11,"summary":"CSAF 2.0 puts the fix, the workaround, the restart cost and the reason a product is not affected in two arrays: remediations and threats. Here is what each field means, how to resolve them per product, and three places where Trivy v0.76.2 and gocsaf v3.7.0 diverge from the spec.","tags":["csaf","csaf vex","vex","trivy","gocsaf","vulnerability management","sbom"],"takeaways":["In CSAF 2.0, remediations and threats each need a category and details, plus product_ids or group_ids to say which products they apply to. A remediation with neither fails mandatory test 6.1.29.","Remediation category has five values in CSAF 2.0: vendor_fix, workaround, mitigation, none_available and no_fix_planned. vendor_fix cannot sit alongside none_available or no_fix_planned for the same product.","Threat category impact is where a CSAF VEX file explains why a known_not_affected product is not affected, if it does not use a flag. Resolve it per product: one vulnerability can carry several impact threats.","Trivy v0.76.2 records the first impact threat in a vulnerability as the statement for every product it filters, whatever that threat's product_ids say, and never reads remediations.","Trivy decodes CSAF with gocsaf v3.7.0, which models product_tree.product_groups as an object, not the array the schema requires. In our test, a schema-valid CSAF 2.0 file with product_groups failed to decode, and Trivy's loader returns that error instead of applying the file."],"title":"CSAF VEX Remediations and Threats: What Trivy Gets Wrong","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-21","url":"https://firewallsync.com/posts/csaf-vex-remediations-threats-category-product-ids-trivy/"},{"author":"FirewallSync Editorial","description":"Why SMS MFA is being phased out — SIM swapping, SS7 interception, and real-time phishing relay — and what NIST and regulators now recommend instead.","faq":[],"markdown_url":"https://firewallsync.com/posts/why-sms-mfa-is-being-replaced-and-what-to-use-instead/index.md","published":"2026-11-20","reading_minutes":5,"summary":"SMS-based MFA doesn't fail because of a flaw in the code itself. It fails because it depends on the telephone network's identity verification, which was never designed to be a security control.","tags":["authentication","iam"],"takeaways":[],"title":"Why SMS MFA Is Being Replaced—and What to Use Instead","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-11-20","url":"https://firewallsync.com/posts/why-sms-mfa-is-being-replaced-and-what-to-use-instead/"},{"author":"Craig Leslie","description":"Chase, Visa and Mastercard say vendors were breached and some names and emails may be exposed, not card data. What each firm said and what is unconfirmed.","faq":[{"a":"None of the three companies says so. Visa said no payments data such as card numbers was affected, Mastercard said the exposed data was non-payment email addresses, and JPMorgan Chase said the data may include some customer names and email addresses. JPMorgan said it will contact affected customers directly where appropriate.","q":"Was my Chase, Visa or Mastercard card number stolen?"},{"a":"Not disclosed. None of the three firms named its third-party provider. Visa said its provider is in Asia Pacific. It is also not confirmed whether the three incidents involve the same vendor.","q":"Which vendor was hacked in the Chase, Visa and Mastercard breach?"},{"a":"An extortion group called n0n listed all three companies on its data-leak site, according to The Register, and SOCRadar told Yahoo Finance the group shared sample data. SOCRadar said it had not independently verified that the group carried out the intrusions. No company has attributed the incident.","q":"Who is behind the attack?"},{"a":"JPMorgan Chase and Visa both said no. Chase said there is no connection to service disruption some customers experienced earlier in the week, and Visa said the incident is unrelated to system disruptions across industries that week.","q":"Is this linked to the bank outages earlier in the week?"}],"markdown_url":"https://firewallsync.com/posts/chase-visa-mastercard-third-party-vendor-breach-what-we-know/index.md","published":"2026-11-20","reading_minutes":6,"summary":"JPMorgan Chase, Visa and Mastercard each said on November 19, 2026 that a third-party provider was hit and that limited non-payment data, such as names and email addresses, may have been exposed. None named the vendors. Here is what the companies said, what the n0n extortion group claims, and what is not confirmed.","tags":["hacking news","third-party risk","data breach","jpmorgan chase","visa","mastercard","phishing","vendor security"],"takeaways":["On November 19, 2026, JPMorgan Chase, Visa and Mastercard each said a third-party service provider had suffered a security incident. None named the provider.","JPMorgan Chase said it believes attackers accessed marketing data held by vendors that may include some customer names and email addresses. Mastercard said a limited number of email addresses of cardholders in one benefit program were exposed.","All three said their own systems and payment data were not affected. Visa said its provider is in Asia Pacific and that card numbers were not affected.","The extortion group n0n claimed all three on its leak site, according to The Register and SOCRadar. SOCRadar told Yahoo Finance it had not independently verified the claim.","The practical risk the companies' statements point to is targeted phishing that uses real names and email addresses. For engineers, the lesson is how much customer data marketing vendors hold."],"title":"Chase, Visa, Mastercard Vendor Breach: What We Know","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-20","url":"https://firewallsync.com/posts/chase-visa-mastercard-third-party-vendor-breach-what-we-know/"},{"author":"John Marshall","description":"OpenVEX statement fields Trivy acts on: status and justification only. status_notes, action_statement, last_updated are ignored; version and supplier dropped.","faq":[{"a":"In Trivy v0.76.2, the products, subcomponents and vulnerability name decide whether a statement matches. Of the matching statements, Trivy takes the one with the newest timestamp and uses its status and justification. It suppresses the finding only for not_affected or fixed, and records the justification as the Statement of the modified finding.","q":"Which OpenVEX fields does Trivy use?"},{"a":"Not in go-vex v0.2.9 or Trivy. SortStatements compares the statement timestamp, or the document timestamp when a statement has none. last_updated is decoded and written back, but not compared. To supersede a statement, add a new statement with a newer timestamp.","q":"Does last_updated change which OpenVEX statement wins?"},{"a":"The OpenVEX spec defines an optional statement-level version, but the go-vex v0.2.9 Statement struct has no field for it. Go's JSON decoder discards unknown keys, so vexctl add, vexctl merge and other go-vex based tools write the document back without it. Statement supplier is dropped the same way.","q":"Why did my OpenVEX statement version disappear?"},{"a":"Yes, by the spec and the JSON schema: an affected statement MUST include one. go-vex's Statement.Validate rejects an affected statement without it, but Trivy does not call that check when it loads a VEX file. vexctl fills the field with the placeholder \"No action statement provided\" when you omit --action-statement.","q":"Is action_statement required for affected status?"}],"markdown_url":"https://firewallsync.com/posts/openvex-statement-fields-status-notes-action-statement-version-trivy/index.md","published":"2026-11-20","reading_minutes":10,"summary":"OpenVEX v0.2.0 defines 13 statement fields, but Trivy v0.76.2 acts on two of them: status and justification, from the matching statement with the newest timestamp. Here is what happens to status_notes, action_statement, last_updated, version and supplier, read from the go-vex v0.2.9 and Trivy source.","tags":["vex","openvex","trivy","go-vex","vexctl","vulnerability management","sbom"],"takeaways":["Trivy v0.76.2 reads two fields from an OpenVEX statement: status (it suppresses only not_affected and fixed) and justification, which it copies into the Statement of a modified finding.","status_notes, impact_statement, action_statement, action_statement_timestamp and last_updated are decoded by go-vex v0.2.9 but play no part in Trivy's decision.","When several statements match, go-vex sorts them by timestamp only, falling back to the document timestamp. Bumping last_updated or a statement version does not make a statement win.","go-vex v0.2.9 has no struct field for statement version or supplier, so tools built on it, such as vexctl add and merge, drop both when they rewrite a document.","Trivy does not run go-vex's statement validation, so an affected statement without an action_statement, which the spec says MUST have one, still loads."],"title":"OpenVEX Statement Fields: What Trivy and go-vex Ignore","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-20","url":"https://firewallsync.com/posts/openvex-statement-fields-status-notes-action-statement-version-trivy/"},{"author":"FirewallSync Editorial","description":"Passkeys vs FIDO2 security keys: both are phishing-resistant by the same underlying protocol, but they differ in storage, portability, and rollout tradeoffs.","faq":[],"markdown_url":"https://firewallsync.com/posts/phishing-resistant-mfa-passkeys-vs-fido2-security-keys/index.md","published":"2026-11-19","reading_minutes":4,"summary":"Passkeys and FIDO2 security keys share the same underlying cryptographic protocol and the same phishing resistance. Where they differ is storage, portability, and what that means for a rollout.","tags":["authentication","iam"],"takeaways":[],"title":"Phishing-Resistant MFA: Passkeys vs FIDO2 Security Keys","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-11-19","url":"https://firewallsync.com/posts/phishing-resistant-mfa-passkeys-vs-fido2-security-keys/"},{"author":"John Marshall","description":"agent-framework 1.26.1 and 1.26.2 on PyPI carried a credential stealer that ran on every Python start. Timeline, IOCs, what is unconfirmed and how to check.","faq":[{"a":"Versions 1.26.1 and 1.26.2 of the agent-framework package on PyPI, both published on November 17, 2026. Wiz names 1.26.0, published November 12, as the most recent safe version, and PyPI lists 1.26.0 as the latest release.","q":"Which agent-framework versions are malicious?"},{"a":"Possibly. Both releases installed a .pth file into site-packages, and Python runs import lines in .pth files every time the interpreter starts. Any Python process in an environment where 1.26.1 or 1.26.2 was installed could run the stealer.","q":"Was I affected if I never imported agent_framework?"},{"a":"No. Starting Python in the affected environment can run the payload again, so Chainguard advises removing the files with shell tools first. Because the stealer sends data out on first run, you also need to rotate every credential the environment could read.","q":"Is uninstalling the package with pip enough?"},{"a":"No one has been publicly attributed as of November 19, 2026. SecurityWeek reports that the publishing method remains unknown.","q":"Who was behind the agent-framework compromise?"}],"markdown_url":"https://firewallsync.com/posts/microsoft-agent-framework-pypi-compromise-1-26-1-1-26-2-what-we-know/index.md","published":"2026-11-19","reading_minutes":7,"summary":"Two malicious releases of Microsoft's agent-framework Python package, 1.26.1 and 1.26.2, were on PyPI for under an hour on November 17, 2026. Both installed a .pth file that ran a credential stealer on every Python start. What is confirmed, what is not, and how to check your environments.","tags":["hacking news","supply chain","pypi","python","credential theft","github actions","azure","ai agents"],"takeaways":["Two releases of Microsoft's agent-framework package on PyPI, 1.26.1 and 1.26.2, were published on November 17, 2026 and were gone by about 21:30 UTC, according to Wiz. PyPI now lists 1.26.0 (November 12) as the latest release.","Both releases dropped a .pth file into site-packages. Python executes import lines in .pth files at interpreter start, so the stealer ran on any Python start in that environment, even if nothing imported agent_framework.","Researchers say the stealer collected cloud, CI and package-registry secrets, including GitHub Actions and Azure DevOps OIDC tokens and Azure managed-identity tokens, and posted them to a webhook.site URL.","How the attacker got publishing access is not confirmed. Wiz says it appears a client certificate and private key for Microsoft's ESRP publishing service were compromised.","If either version was installed anywhere, remove the payload files without starting that Python, rotate every secret the environment could read, and review Azure and GitHub federated identities."],"title":"Microsoft agent-framework PyPI Compromise: What We Know","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-19","url":"https://firewallsync.com/posts/microsoft-agent-framework-pypi-compromise-1-26-1-1-26-2-what-we-know/"},{"author":"Craig Leslie","description":"SSVC Exploitation poc flags 47,926 CVEs with a public PoC in CISA-ADP data. Pull them to CSV, slice by Automatable and Technical Impact, and join EPSS.","faq":[{"a":"It is CISA's assessment at analysis time that a public proof of concept exists or the flaw has a well-known exploitation method. The current SSVC definition (Exploitation 1.1.0) reads: \"Typical public PoC exists in sources such as Metasploit or websites like ExploitDB; or (2) the vulnerability has a well-known method of exploitation.\" It does not mean exploitation in the wild; that is active.","q":"What does Exploitation poc mean in a CISA-ADP SSVC block?"},{"a":"In the cvelistV5 snapshot of 2026-11-18 00:00 UTC, 47,926 CVE records carry a CISA-ADP SSVC block with Exploitation poc. The number changes daily as CISA scores new records and moves some to active.","q":"How many CVEs have a public PoC according to CISA?"},{"a":"When a CVE reaches KEV, CISA-ADP moves its Exploitation value to active, so the current poc set is almost entirely non-KEV. In our snapshot only one poc record was in KEV, and it had been added to KEV that day. To find poc CVEs that later became KEV entries you need history, for example a diff of cvelistV5 snapshots.","q":"Do poc CVEs end up in the KEV catalog?"},{"a":"Not on that signal alone. 47,926 records is too many to treat as urgent, and most score low in EPSS. Combine poc with Automatable, Technical Impact, your exposure and EPSS, and put KEV entries ahead of all of them.","q":"Should I patch every CVE with a public PoC first?"}],"markdown_url":"https://firewallsync.com/posts/public-poc-cves-cisa-adp-ssvc-exploitation-poc-epss/index.md","published":"2026-11-19","reading_minutes":9,"summary":"CISA-ADP marks 47,926 CVE records with SSVC Exploitation poc, meaning a public proof of concept or a well-known exploitation method. We pulled them from the cvelistV5 corpus, joined EPSS and KEV, and show what the label does and does not tell you, with a script that writes the full list to CSV.","tags":["vulnerability management","ssvc","cisa vulnrichment","cisa-adp","epss","kev","cvelistv5","proof of concept"],"takeaways":["In the cvelistV5 snapshot of 2026-11-18 00:00 UTC, 222,849 CVE records carry a CISA-ADP SSVC block. 47,926 (21.5%) have Exploitation poc, 1,787 have active and 173,136 have none.","poc is not a staging area for KEV. Only one poc record is in KEV, because CISA moves a record to active when it reaches the catalog: 1,782 of the 1,786 KEV entries read active, and the other four were added on the day of the snapshot.","Most poc CVEs score low in EPSS: the median is 0.00478 on the 2026-11-18 scores, and 1,231 (2.6%) score 0.1 or higher. A public PoC is a capability signal, not a forecast.","Slice poc by the other two CISA-ADP decision points. 5,328 poc records are also Automatable yes and Technical Impact total; that slice has roughly twice the median EPSS of the full poc set.","The script in this article writes every poc CVE with its Automatable, Technical Impact and EPSS values to CSV in about 40 seconds from the cvelistV5 release zip."],"title":"Public PoC CVEs: Pull All 47,926 SSVC poc Records","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-19","url":"https://firewallsync.com/posts/public-poc-cves-cisa-adp-ssvc-exploitation-poc-epss/"},{"author":"FirewallSync Editorial","description":"How to build an AI security policy developers will actually follow: risk-tiered rules, fast approval paths, and concrete guidance instead of blanket restrictions.","faq":[],"markdown_url":"https://firewallsync.com/posts/how-to-build-an-ai-security-policy-without-blocking-developers/index.md","published":"2026-11-18","reading_minutes":5,"summary":"A policy that blanket-bans AI tools gets quietly worked around. A policy that gives developers a fast, clear path to use them safely actually gets followed.","tags":["ai security","process"],"takeaways":[],"title":"How to Build an AI Security Policy Without Blocking Developers","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-18","url":"https://firewallsync.com/posts/how-to-build-an-ai-security-policy-without-blocking-developers/"},{"author":"Craig Leslie","description":"Equinor says an ongoing cyberattack hit certain IT systems on Nov 17. Production unaffected, no sign of data theft so far. Confirmed vs unknown, and lessons.","faq":[{"a":"Equinor says it detected an ongoing cyberattack against certain IT systems on November 17, 2026. It has not said how the attackers got in or how far they reached, and it says there are no indications that data has been stolen.","q":"Was Equinor hacked?"},{"a":"According to Equinor's statement, no: 'The attack has not affected Equinor's operations and production.' Reuters reported the same. The company describes the affected systems only as 'certain IT systems.'","q":"Did the Equinor cyberattack affect oil and gas production?"},{"a":"Equinor has not identified the attackers, and in the reporting we reviewed on November 18, 2026 no group had publicly claimed responsibility. Treat any attribution you see as unconfirmed until Equinor or Norwegian authorities state it.","q":"Who is behind the Equinor attack?"},{"a":"Equinor says 'There are no indications that data has been stolen.' That is the company's finding at the time of the statement while the attack was described as ongoing; it can change as the investigation continues.","q":"Was Equinor customer or employee data stolen?"}],"markdown_url":"https://firewallsync.com/posts/equinor-cyberattack-it-systems-what-we-know/index.md","published":"2026-11-18","reading_minutes":5,"summary":"Norwegian energy company Equinor said on Tuesday, November 17, 2026 that it has detected an ongoing cyberattack against certain IT systems. It says operations and production are unaffected, there are no indications data was stolen, and authorities are being informed. Here is what is confirmed, what is not, and what engineers can take from it.","tags":["hacking news","equinor","energy sector","incident response","breach disclosure","norway","nsm"],"takeaways":["Equinor published a statement on November 17, 2026 (18:55 CET) saying it 'has detected an ongoing cyber-attack against certain IT systems.'","Equinor says the attack 'has not affected Equinor's operations and production' and that 'There are no indications that data has been stolen.'","The company says relevant resources have been mobilised 'to neutralise the attack and regain full integrity' and that relevant authorities are being informed.","Equinor has not said which systems are involved, how the attackers got in, whether ransomware is involved or who is responsible. No group has publicly claimed the attack in the reporting we reviewed.","The disclosure comes weeks after Norway's National Security Authority (NSM) raised the national cyber risk level to HIGH, the second-highest step on its five-level scale. Nothing public links the two."],"title":"Equinor Cyberattack: What the Energy Giant Has Confirmed","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-18","url":"https://firewallsync.com/posts/equinor-cyberattack-it-systems-what-we-know/"},{"author":"John Marshall","description":"KEV dueDate is now 3 or 14 days after dateAdded; it was 21 days from 2022 to early 2026. Counts by year, forensicTriage link, edited due dates, and a script.","faq":[{"a":"The catalog's dueDate sets it per entry. In KEV catalog version 2026.11.17, every entry added since June 10, 2026 is due either 3 or 14 calendar days after dateAdded. Between March 2022 and early March 2026 the usual gap was 21 days. For federal civilian agencies the binding timeline comes from BOD 26-04, which can differ per asset; the catalog shows one date per CVE.","q":"How many days does CISA give to fix a KEV vulnerability?"},{"a":"In the current catalog, yes: all 105 entries added since June 10, 2026 with forensicTriage Yes have a dueDate 3 days after dateAdded. The reverse is not true. 23 entries with forensicTriage No also have a 3-day gap, so use dueDate rather than inferring the deadline from forensicTriage.","q":"Does forensicTriage Yes mean a 3-day due date?"},{"a":"Yes. Replaying every catalog version in the cisagov/kev-data repository since January 27, 2025 shows 23 dueDate changes. Some fixed data errors, some shortened a deadline (for example CVE-2025-0108 from April 8 to March 11, 2025) and one lengthened it (CVE-2026-86950, when its forensicTriage value changed from Yes to No).","q":"Can a KEV due date change after the entry is published?"},{"a":"No. The gap is a whole number of calendar days. Of the 165 entries added since June 10, 2026, 49 have a dueDate that falls on a Saturday or Sunday, so a ticketing SLA that counts business days will run past the catalog date.","q":"Are KEV due dates business days?"}],"markdown_url":"https://firewallsync.com/posts/kev-duedate-days-after-dateadded-3-14-21-day-deadlines/index.md","published":"2026-11-18","reading_minutes":8,"summary":"We measured dueDate minus dateAdded for all 1,782 entries in CISA's KEV catalog (version 2026.11.17) and replayed the cisagov/kev-data history to find every dueDate CISA edited after publication. The default gap was 14 days in 2021, 21 days from March 2022, and since mid-2026 is 3 or 14 days. Every forensicTriage Yes entry is due in 3 days, but 23 No entries are too.","tags":["vulnerability management","kev","cisa","bod 26-04","bod 22-01","forensic triage","kev-data","sla"],"takeaways":["In KEV catalog version 2026.11.17 (1,782 entries), the gap between dateAdded and dueDate is 3 or 14 days for every entry added since June 10, 2026: 128 entries at 3 days and 37 at 14 days.","From March 2022 to early March 2026 the catalog's usual gap was 21 days, not the two weeks BOD 22-01 set as its default. 2021 entries used 14 days for newer CVEs and about six months (181 or 182 days) for CVE IDs from before 2021.","All 105 entries added since June 10, 2026 with forensicTriage Yes are due in 3 days. 23 entries with forensicTriage No are also due in 3 days, so the field does not tell you the deadline: read dueDate.","The CISA-ADP SSVC Technical Impact value in the CVE record does not predict dueDate either: 19 entries with Technical Impact total were given 14 days.","CISA edits published due dates. The kev-data history since January 2025 shows 23 dueDate changes, including shortened deadlines; one (CVE-2022-0995) was moved to a date that had already passed. Re-read dueDate on every sync instead of caching it."],"title":"KEV dueDate: 3, 14 or 21 Days After dateAdded, 2021-2026","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-18","url":"https://firewallsync.com/posts/kev-duedate-days-after-dateadded-3-14-21-day-deadlines/"},{"author":"FirewallSync Editorial","description":"Shadow AI explained: how employees pasting company data into unapproved AI tools creates real exposure, and why blocking access doesn't actually solve it.","faq":[],"markdown_url":"https://firewallsync.com/posts/shadow-ai-how-employees-accidentally-create-security-gaps-with-ai-tools/index.md","published":"2026-11-17","reading_minutes":5,"summary":"Shadow AI isn't employees deliberately bypassing security. It's people solving real work problems with the tools available to them, without a sanctioned option existing yet — and the data flows that follow are usually invisible until something goes wrong.","tags":["ai security","process"],"takeaways":[],"title":"Shadow AI: How Employees Accidentally Create Security Gaps With AI Tools","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-17","url":"https://firewallsync.com/posts/shadow-ai-how-employees-accidentally-create-security-gaps-with-ai-tools/"},{"author":"John Marshall","description":"Coles told the ASX on November 16 it is investigating a cyber incident. Stores and online run normally; no evidence so far of data theft. Confirmed vs unknown.","faq":[{"a":"Coles says it is investigating a cyber security incident affecting its IT environment and is working to contain it. It has not described the incident in more detail, and it says it has found no evidence so far that customer, team or supplier data was compromised or removed.","q":"Was Coles hacked?"},{"a":"Not according to Coles' initial findings. Its ASX update, as reported on November 16, 2026, says it has found no evidence so far that customer, team or supplier data was compromised or removed. The investigation is continuing, so that could change.","q":"Is Coles customer data affected?"},{"a":"Yes, according to Coles: \"Operations continue and customers are able to shop as normal with Coles in stores and online.\" Information Age separately reported that a small number of customers had problems with Liquorland click-and-collect orders; Coles has not linked that to the incident.","q":"Can I still shop at Coles?"},{"a":"Nobody has been identified. Coles has not named a threat actor, and none of the reporting we reviewed on November 16 and 17, 2026 mentions a public claim of responsibility.","q":"Who is behind the Coles cyber attack?"}],"markdown_url":"https://firewallsync.com/posts/coles-cyber-security-incident-what-we-know/index.md","published":"2026-11-17","reading_minutes":4,"summary":"Australian supermarket group Coles told the ASX on Monday, November 16, 2026 that it is investigating a cyber security incident in its IT environment. It says stores and online shopping are running normally and it has found no evidence so far that customer, team or supplier data was compromised or removed. Here is what is confirmed, what is not, and what engineers can take from the disclosure.","tags":["hacking news","coles","retail security","incident response","breach disclosure","australia","acsc"],"takeaways":["Coles disclosed a cyber security incident to the ASX on Monday, November 16, 2026, according to Forbes Australia, Information Age and Inside Retail.","Coles says operations continue and customers can shop as normal in stores and online, and that it has found no evidence so far that customer, team or supplier data was compromised or removed.","Coles says it is working with the Australian Cyber Security Centre and global cyber security experts to contain and secure its IT environment.","Coles has not said what kind of incident it is, which systems are involved or when it started. No group has publicly claimed it in the reporting we reviewed.","\"No evidence so far\" is an interim finding, not a clean bill of health; suppliers and staff should expect follow-up notices and watch for phishing that uses the news."],"title":"Coles Cyber Incident: What the Supermarket Has Confirmed","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-17","url":"https://firewallsync.com/posts/coles-cyber-security-incident-what-we-know/"},{"author":"Craig Leslie","description":"KEV in CVE records: 1,777 of 1,781 KEV CVEs carry a matching CISA-ADP kev block. 2 lack one, 2 have a different dateAdded, removed CVEs lose theirs. Script.","faq":[{"a":"Almost. In our November 17, 2026 check, 1,779 of the 1,781 CVEs in KEV catalog version 2026.11.13 had a CISA-ADP kev block in their cvelistV5 record. The two exceptions, CVE-2026-48356 and CVE-2026-111497, were added to KEV on 2026-10-27 and had an SSVC Exploitation value of active but no kev block.","q":"Does every KEV vulnerability have a kev block in its CVE record?"},{"a":"For the three removals visible in the cisagov/kev-data history since January 2025 (CVE-2025-4664, CVE-2025-6264 and CVE-2026-69836), the current CVE records have no kev block and their CISA-ADP SSVC Exploitation value is none. A scan of all 426,227 records found no kev block for any CVE outside the catalog.","q":"What happens to a CVE record when CISA removes a vulnerability from KEV?"},{"a":"Use the catalog JSON as the primary source: it is what CISA publishes first, and it carries dueDate, requiredAction and knownRansomwareCampaignUse, which the kev block does not. The kev block is useful when your pipeline already ingests CVE records, provided you reconcile it against the catalog and treat the catalog's dateAdded as authoritative.","q":"Should I use the kev block in the CVE record or the KEV catalog JSON?"},{"a":"Download known_exploited_vulnerabilities.json from cisagov/kev-data, fetch each listed CVE's JSON from CVEProject/cvelistV5, and run the Python script in this article. It prints the CVEs with no kev block and those whose dateAdded differs from the catalog. On November 17, 2026 the fetch of 1,781 records took about 11 seconds with parallel curl.","q":"How do I run this audit myself?"}],"markdown_url":"https://firewallsync.com/posts/kev-in-cve-records-cisa-adp-kev-block-full-catalog-audit/index.md","published":"2026-11-17","reading_minutes":8,"summary":"We compared every entry in CISA's KEV catalog (version 2026.11.13, 1,781 CVEs) with the CISA-ADP kev block in each CVE record, and scanned all 426,227 records in the November 17, 2026 cvelistV5 snapshot for kev blocks that point at CVEs no longer in the catalog. The record copy matches the catalog for 1,777 CVEs; this article lists the four that don't, what happens to a record when CISA removes a CVE from KEV, and a script to run the check yourself.","tags":["vulnerability management","kev","cisa-adp","vulnrichment","cve json 5","cvelistv5","ssvc","data quality"],"takeaways":["Across the whole KEV catalog (version 2026.11.13, 1,781 CVEs), 1,779 CVE records carry a CISA-ADP kev block and 1,777 of those blocks have the same dateAdded as the catalog.","Two KEV CVEs added on 2026-10-27 (Adobe Commerce CVE-2026-48356 and Google Brotli CVE-2026-111497) still had no kev block in their records on November 17, although their CISA-ADP SSVC Exploitation value is active.","Two PaperCut NG/MF CVEs (CVE-2026-81578 and CVE-2026-82078) carry a kev dateAdded of 2026-08-28 in the record while the catalog says 2026-08-31.","No CVE record in the November 17 corpus (426,227 records) carries a kev block for a CVE that is not in the catalog. The three CVEs that the kev-data history shows were removed since January 2025 now have no kev block and an SSVC Exploitation value of none.","If you key exploitation decisions off CVE records, the record is a near-complete mirror of KEV, but reconcile it against the catalog JSON on a schedule. The catalog remains the source of truth for dateAdded, dueDate and ransomware use."],"title":"KEV in CVE Records: Audit of All 1,781 CISA-ADP kev Blocks","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-17","url":"https://firewallsync.com/posts/kev-in-cve-records-cisa-adp-kev-block-full-catalog-audit/"},{"author":"FirewallSync Editorial","description":"A production checklist for securing MCP servers: authentication, credential scoping, input validation, and monitoring for the specific attack patterns MCP introduces.","faq":[],"markdown_url":"https://firewallsync.com/posts/how-to-secure-mcp-servers-in-production/index.md","published":"2026-11-16","reading_minutes":5,"summary":"Running an MCP server in production means running a piece of infrastructure that mediates between an AI model and real systems — it deserves the same operational rigor as any other production service handling credentials.","tags":["ai security"],"takeaways":[],"title":"How to Secure MCP Servers in Production","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-16","url":"https://firewallsync.com/posts/how-to-secure-mcp-servers-in-production/"},{"author":"Craig Leslie","description":"La Poste says Maileva lost names and postal addresses of 22 million people in September. Bank details not affected. Confirmed facts, timeline and checks.","faq":[{"a":"According to La Poste, quoted by AFP via Euronews, the exposed data is the first name, last name and postal address of the people concerned. La Poste says other personal information such as bank details is not affected. The people concerned are recipients of mail sent by Maileva's business customers, not necessarily Maileva account holders.","q":"What data was stolen in the Maileva breach?"},{"a":"22 million people, according to La Poste and to a written response from Maileva published by Cyberattaque.org. Cyberattaque.org also reports that internal documents it attributes to Docaposte list about 39 million references that deduplicate to roughly 22 million people; that document figure has not been confirmed publicly by the company.","q":"How many people are affected by the Maileva breach?"},{"a":"Maileva told Cyberattaque.org that all affected organizations were informed of the incident and that informing the people concerned is the responsibility of Maileva's client organizations. That matches the GDPR split in which a processor notifies the controller and the controller handles notification.","q":"Who will notify people affected by the Maileva breach?"},{"a":"Nobody has been named. None of the sources we reviewed reports a claim of responsibility or an attribution, and the method of access has not been disclosed.","q":"Who is behind the Maileva attack?"}],"markdown_url":"https://firewallsync.com/posts/maileva-data-breach-la-poste-22-million-addresses-what-we-know/index.md","published":"2026-11-16","reading_minutes":6,"summary":"La Poste told AFP on November 14, 2026 that its subsidiary Maileva, a business mail-sending platform, lost the first names, last names and postal addresses of 22 million people in a mid-September attack. What is confirmed, what is not, and what teams that upload recipient lists to a mailing platform should check.","tags":["hacking news","maileva","la poste","docaposte","data breach","gdpr","saas security","third-party risk"],"takeaways":["La Poste confirmed to AFP on November 14, 2026 that Maileva, a Docaposte subsidiary that sends mail on behalf of businesses, suffered a mid-September theft of the contact details of 22 million people.","The exposed data is first name, last name and postal address of recipients of Maileva's business customers. La Poste says bank details and other personal information are not affected.","Maileva suspended all services on September 23 after what it described as unauthorized access attempts to client accounts aimed at sending mail illegitimately; Docaposte said services were fully back on September 29.","Maileva notified France's data protection authority CNIL and filed a criminal complaint; a judicial investigation is under way. Telling affected people is left to Maileva's business customers, who are the data controllers.","No attacker has been publicly named, and how the attacker got in has not been disclosed."],"title":"Maileva Data Breach: La Poste Says 22 Million Hit","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-11-16","url":"https://firewallsync.com/posts/maileva-data-breach-la-poste-22-million-addresses-what-we-know/"},{"author":"John Marshall","description":"Alert on CVE exploitation by diffing CISA-ADP kev and SSVC Exploitation. 30 days of cvelistV5 data: every new active was a KEV CVE; 2 KEV adds lacked a block.","faq":[{"a":"In the CISA-ADP container: containers.adp[] where providerMetadata.shortName is CISA-ADP. Its metrics[].other entries hold an ssvc object whose content.options include an Exploitation value, and, for CVEs on the KEV catalog, a kev object with content.dateAdded and a reference URL. The CNA container has no standard exploitation field.","q":"Where does a CVE record say a vulnerability is exploited?"},{"a":"Not by definition: SSVC defines active as shared, observable, reliable evidence of in-the-wild use with credible public reporting, while KEV has its own listing criteria. In our 30-day diff, every record that gained or was published with active was a KEV CVE, but that is one month of data, not a rule.","q":"Is SSVC Exploitation active the same as being on CISA KEV?"},{"a":"For 31 KEV additions between October 17 and November 15, 2026 that had a kev block by November 16, the CISA-ADP container's dateUpdated was one day after the KEV dateAdded for 28, two days for 2 and four days for 1. dateUpdated is the container's last update, so it bounds when the block appeared rather than recording it exactly.","q":"How quickly does the KEV block appear in the CVE record?"},{"a":"Alert on the KEV JSON feed first, because it is the source the ADP kev block copies and it was never behind in our data. Diff CISA-ADP as well if you want poc changes, SSVC values for CVEs that are not on KEV, and REJECTED state changes in the same pipeline.","q":"Should I alert on the KEV feed or on CVE records?"}],"markdown_url":"https://firewallsync.com/posts/cve-exploitation-alerts-cisa-adp-kev-ssvc-cvelistv5-diff/index.md","published":"2026-11-16","reading_minutes":8,"summary":"A CVE record's CISA-ADP container gains a KEV block and an SSVC Exploitation value of active when exploitation is confirmed. We diffed 31,312 changed records between the October 17 and November 16, 2026 cvelistV5 snapshots to see how those signals move, then built a small watcher that alerts on them.","tags":["vulnerability management","cisa-adp","vulnrichment","ssvc","kev","cvelistv5","alerting","cve json 5"],"takeaways":["CISA-ADP puts two exploitation signals in a CVE record: an SSVC block with an Exploitation value (none, poc or active) and, for CVEs on the KEV catalog, a kev block with dateAdded. Both sit in containers.adp[].metrics[].other.","Between the cvelistV5 snapshots of October 17 and November 16, 2026, 34 CVE records gained or were published with Exploitation active (16 changed, 18 new). All 34 were on the KEV catalog, so in this window active added no CVE that KEV did not already list.","Of 33 CVEs CISA added to KEV in the window, 31 carried a CISA-ADP kev block by November 16. For 28 of them the CISA-ADP container was last updated the day after the KEV dateAdded; the other three took two or four days. Two (Adobe CVE-2026-48356 and Google CVE-2026-111497) had Exploitation active but no kev block.","Exploitation also moved to poc on 292 changed records and occasionally back down from poc to none (5 records), so treat it as a current state, not a ratchet.","Use the KEV JSON feed as the primary exploitation alert and the CISA-ADP diff as a second signal that also catches poc changes and rejections. The watcher script in this article does the diff from cvelistV5 delta.json."],"title":"CVE Exploitation Alerts: Diff CISA-ADP KEV and SSVC Data","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-16","url":"https://firewallsync.com/posts/cve-exploitation-alerts-cisa-adp-kev-ssvc-cvelistv5-diff/"},{"author":"FirewallSync Editorial","description":"How MCP authentication and authorization actually work under OAuth 2.1: audience binding, the confused deputy problem, and what a secure remote MCP deployment must validate.","faq":[],"markdown_url":"https://firewallsync.com/posts/mcp-authentication-and-authorization-what-a-secure-deployment-should-check/index.md","published":"2026-11-15","reading_minutes":5,"summary":"MCP's authorization spec is built on OAuth 2.1 for a specific reason: most MCP clients are desktop apps and CLI tools that can't securely hold a client secret, and OAuth 2.1's mandatory PKCE was designed for exactly that constraint.","tags":["ai security","authentication"],"takeaways":[],"title":"MCP Authentication and Authorization: What a Secure Deployment Should Check","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-15","url":"https://firewallsync.com/posts/mcp-authentication-and-authorization-what-a-secure-deployment-should-check/"},{"author":"John Marshall","description":"Huntress counted 288 hosts in 48 orgs, then 205 in 41, hit by rogue ScreenConnect clients tied to the RMM+ extension. Confirmed facts, timeline and checks.","faq":[{"a":"A campaign Huntress first observed on November 11, 2026 in which rogue ScreenConnect clients were silently installed on managed endpoints, followed by a roughly 16-file toolkit Huntress calls SCKit. Huntress counted 288 hosts across 48 organizations in the first wave and 205 hosts across 41 organizations in a second wave on November 13.","q":"What is the ScreenConnect SCKit attack?"},{"a":"ConnectWise said it found no evidence of a compromise of ConnectWise infrastructure or a vulnerability in ScreenConnect itself, and that its initial assessment points to abuse of independent third-party extensions. Separately, it told partners to confirm technician clients run 26.6.5 or later, the release that fixes CVE-2026-84869, a client-side file-transfer flaw that CISA added to KEV on September 11, 2026.","q":"Is ScreenConnect itself vulnerable?"},{"a":"Labtech Plus, RMM Plus, RMM+, RMM Plus Password Link and RMM+ Password Link, across about 400 partner environments where it could act remotely. It asked self-hosted (on-premises) partners to disable them as well. ConnectWise noted the roughly 400 environments were flagged for having the extensions installed, not for confirmed compromise.","q":"Which ScreenConnect extensions did ConnectWise disable?"},{"a":"Neither Huntress nor ConnectWise has attributed it. Huntress notes the attacker participant name 'Addmin' in session data but does not link it to a known group.","q":"Who is behind the SCKit campaign?"}],"markdown_url":"https://firewallsync.com/posts/screenconnect-sckit-worm-rmm-plus-extension-what-we-know/index.md","published":"2026-11-15","reading_minutes":5,"summary":"Huntress reports rogue ScreenConnect clients and an SCKit toolkit on 288 hosts at 48 organizations on November 11, 2026 and 205 more hosts at 41 organizations on November 13. ConnectWise disabled five third-party extensions, including RMM+, across about 400 partner environments. What is confirmed, what is not, and what to check.","tags":["hacking news","screenconnect","connectwise","rmm","msp security","sckit","incident response"],"takeaways":["Huntress says rogue ScreenConnect clients and a toolkit it calls SCKit reached at least 288 managed endpoints across 48 organizations in about 90 minutes on November 11, 2026, and a second wave hit 205 hosts across 41 organizations on November 13.","ConnectWise said it found no evidence of a compromise of its infrastructure or a vulnerability in ScreenConnect itself, and disabled five independently developed extensions (Labtech Plus, RMM Plus, RMM+, RMM Plus Password Link, RMM+ Password Link) across about 400 partner environments.","Huntress describes the link to RMM+ as suggestive, not confirmed: versions 1.0.31 and 1.0.32 accept a blank access key when an administrator never set one. ConnectWise had not identified the initial access vector at the time of its advisory.","If you run ScreenConnect: disable the RMM+-family extensions, confirm technician clients run 26.6.5 or later (the CVE-2026-84869 fix), and do not triage suspect hosts through ScreenConnect itself.","No actor has been named by Huntress or ConnectWise."],"title":"ScreenConnect SCKit Worm via RMM+ Extension: What We Know","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-15","url":"https://firewallsync.com/posts/screenconnect-sckit-worm-rmm-plus-extension-what-we-know/"},{"author":"Craig Leslie","description":"CVE JSON stores CWE in problemTypes descriptions cweId, but 24% of 2026 records have no CNA cweId. Gap sources, CISA-ADP fills, and a parser for every case.","faq":[{"a":"In containers.cna.problemTypes, an array of objects, each with a descriptions array. Each description has lang and description (required) and optional cweId, type and references. The machine-readable CWE is the cweId string, for example CWE-79. ADP containers such as CISA-ADP can carry their own problemTypes with the same structure.","q":"Where is the CWE in a CVE JSON 5 record?"},{"a":"The schema does not require one. CNA Rules 4.1.0 say a record MUST identify the type of vulnerability but only SHOULD use CWE, so a CNA can meet the rule with text such as 'Elevation of privilege' or 'n/a'. Some CNAs omit problemTypes entirely; in the November 15, 2026 snapshot that includes every 2026 record from the Linux kernel CNA.","q":"Why does a CVE record have no CWE?"},{"a":"Yes, through the CISA-ADP container (Vulnrichment). CISA says it adds CWE and CVSS for some higher-risk CVEs where the CNA has not, never overwrites the CNA container, and drops its own values if the CNA later adds them. In 2026 records it supplied a cweId for 15,650 records that had none from the CNA.","q":"Does CISA add CWE to CVE records?"},{"a":"Count it, but flag it. CWE lists CWE-284 Improper Access Control as a Pillar with vulnerability mapping usage DISCOURAGED and says it is not useful for trend analysis. In 2026 records it is the most common CWE that CISA-ADP adds when the CNA gave none, so a dashboard that merges sources without labels will inflate it.","q":"Should I count CWE-284 in a weakness dashboard?"}],"markdown_url":"https://firewallsync.com/posts/cve-problemtypes-cweid-cwe-from-cve-json/index.md","published":"2026-11-15","reading_minutes":9,"summary":"The CWE for a CVE lives in problemTypes[].descriptions[].cweId, but a count across the November 15, 2026 cvelistV5 snapshot shows 24% of 2026 CVE records have no CNA cweId. Where the gaps come from, what CISA-ADP fills in (often CWE-284), and a parser that handles all four cases.","tags":["vulnerability management","cwe","cve json 5","problemtypes","cisa-adp","vulnrichment","cvelistv5","parsing"],"takeaways":["In CVE JSON 5.x the CWE sits at containers.cna.problemTypes[].descriptions[].cweId, a string matching ^CWE-[1-9][0-9]*$. cweId is optional, and problemTypes itself is not in the schema's required list for a CNA container.","In the cvelistV5 snapshot of November 15, 2026, 74,189 of 97,709 CVE records published in 2026 (75.9%) carry a CNA cweId. 13,044 (13.3%) have only text entries and 10,476 (10.7%) have no problemTypes at all.","The gaps are concentrated: every 2026 Oracle and Apple record is text-only, no 2026 Linux kernel record includes problemTypes, and WPScan writes the CWE into the description with type CWE but no cweId field.","CISA-ADP adds a cweId to 15,650 of the 2026 records the CNA left without one. Its most common gap-fill is CWE-284, which CWE marks DISCOURAGED for vulnerability mapping.","A robust extractor reads the CNA cweId first, then CWE IDs named in CNA text, then CISA-ADP, and records which source it used so dashboards can separate supplier data from enrichment."],"title":"CVE problemTypes and cweId: Getting CWE From CVE JSON","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-15","url":"https://firewallsync.com/posts/cve-problemtypes-cweid-cwe-from-cve-json/"},{"author":"FirewallSync Editorial","description":"MCP tool poisoning explained: how attackers hide instructions in tool descriptions and parameters, why agents follow them, and how to defend against it.","faq":[],"markdown_url":"https://firewallsync.com/posts/mcp-tool-poisoning-how-malicious-instructions-can-hide-inside-trusted-tools/index.md","published":"2026-11-14","reading_minutes":4,"summary":"An AI agent decides how to use a tool by reading its description. Tool poisoning exploits exactly that: the description isn't documentation to the model, it's an instruction it will follow.","tags":["ai security"],"takeaways":[],"title":"MCP Tool Poisoning: How Malicious Instructions Can Hide Inside Trusted Tools","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-14","url":"https://firewallsync.com/posts/mcp-tool-poisoning-how-malicious-instructions-can-hide-inside-trusted-tools/"},{"author":"Craig Leslie","description":"Uber says compromised credentials at an AI vendor, reported as Fireworks AI, exposed driver data. State filings list 7,870 people so far. What is confirmed.","faq":[{"a":"TechCrunch reports that Uber's notices to drivers say personal information and driver's license information were stolen. Reuters reports that the California filing itself did not list the data involved. Uber says it has not identified any misuse of the data.","q":"What data was stolen in the Uber breach?"},{"a":"Uber's notice says an unauthorized party used compromised credentials at a service provider to access Uber data. TechCrunch identifies the provider as Fireworks AI and reports Uber described it as a security incident impacting one of its third-party vendors. Uber has not said its own systems were breached by other means.","q":"Was Uber itself hacked?"},{"a":"Uber has not given a total. Reported state filings list 2,567 people in Massachusetts, 3,430 in South Carolina and 1,873 in Rhode Island. TechCrunch reports Illinois residents are affected with a count pending and California drivers were notified without a number.","q":"How many Uber drivers were affected?"},{"a":"TechCrunch and Reuters link it to Fireworks AI, which has disclosed two October 2026 incidents: unauthorized use of internal credentials identified on October 6, and an October 23 incident at a cloud provider hosting its R\u0026D machines. Neither Uber nor Fireworks has said publicly which path reached Uber's data.","q":"Is this the same Fireworks AI breach that affected Notion?"}],"markdown_url":"https://firewallsync.com/posts/uber-driver-license-data-stolen-fireworks-ai-breach-what-we-know/index.md","published":"2026-11-14","reading_minutes":5,"summary":"Uber says an unauthorized party used compromised credentials at a service provider to reach Uber data, and TechCrunch reports the provider is Fireworks AI and that drivers' license information was taken. State filings so far list 7,870 people in three states. What is confirmed, the timeline, and what engineering teams should check.","tags":["hacking news","uber","fireworks ai","data breach","ai vendors","third-party risk","credentials"],"takeaways":["Uber's notice on the California attorney general's site, dated November 10, 2026, says an unauthorized party used compromised credentials at a service provider to access Uber data. Reuters reports Uber says it cut its data connections with the provider and has not identified any misuse of the data.","TechCrunch reports the provider is Fireworks AI, the same AI inference vendor whose breach exposed Notion AI chat transcripts, and that drivers' personal and driver's license information was stolen.","Per the reporting, the provider told Uber on October 9 that credentials were compromised, and Uber's investigation found on October 13 that they had been used to access its data.","State filings reported so far list 2,567 people in Massachusetts, 3,430 in South Carolina and 1,873 in Rhode Island (7,870 combined). Illinois and California filings did not give a number in the reporting, so the total is not known.","For engineering teams: inventory every credential your company has handed to an AI vendor (API keys, cloud tokens, data-connector secrets), scope them to the minimum, and plan to revoke them at the issuer the day a vendor reports an incident."],"title":"Uber Driver License Data Stolen in Fireworks AI Breach","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-14","url":"https://firewallsync.com/posts/uber-driver-license-data-stolen-fireworks-ai-breach-what-we-know/"},{"author":"John Marshall","description":"CVSS v4 Supplemental metrics appear in only 3.4% of CNA v4.0 scores in CVE records, mostly from Juniper and Palo Alto. Field names, values and a parser.","faq":[{"a":"No. The CVSS v4.0 specification says no Supplemental metric will have any impact on the final calculated CVSS score. A vector ending in AU:Y/U:Red produces the same CVSS-B or CVSS-BT number as the same vector without those segments.","q":"Do CVSS v4.0 Supplemental metrics change the CVSS score?"},{"a":"In the cvssV4_0 object defined by cve-schema's cvss-v4.0.json they are Safety, Automatable, Recovery, valueDensity, vulnerabilityResponseEffort and providerUrgency. The enums are upper case (for example YES, NO, AMBER, RED) with NOT_DEFINED as the default, while the vector string uses S, AU, R, V, RE and U with short values such as Y, N and Amber.","q":"What are the JSON field names for CVSS v4 Supplemental metrics in a CVE record?"},{"a":"They ask nearly the same question: whether attackers can reliably automate steps 1-4 of the kill chain (reconnaissance, weaponization, delivery, exploitation). They are set by different parties at different times. In CVE-2024-0012 the CNA, Palo Alto Networks, set AU:N in its CVSS vector while CISA's SSVC entry in the same record says Automatable: yes.","q":"Is CVSS Automatable the same as SSVC Automatable?"},{"a":"Use it as one input from the supplier, not as a ranking key. The specification calls Provider Urgency a pass-through metric for the provider's own assessment. Only 1,152 CNA v4.0 entries across all published CVE records set it, so a policy that depends on it will see no value for almost every CVE.","q":"Should I use Provider Urgency to prioritise patches?"}],"markdown_url":"https://firewallsync.com/posts/cvss-v4-supplemental-metrics-in-cve-records-automatable-provider-urgency/index.md","published":"2026-11-14","reading_minutes":9,"summary":"CVSS v4.0 Supplemental metrics (Safety, Automatable, Recovery, Value Density, Response Effort, Provider Urgency) never change the score. A count across every published CVE record shows only 3.4% of CNA v4.0 entries set any of them, which CNAs do, how the JSON fields and vector string drift apart, and how to use them without overtrusting them.","tags":["vulnerability management","cvss","cvss v4","cve json 5","supplemental metrics","automatable","provider urgency","ssvc","parsing"],"takeaways":["CVSS v4.0 defines six optional Supplemental metrics: Safety (S), Automatable (AU), Recovery (R), Value Density (V), Vulnerability Response Effort (RE) and Provider Urgency (U). The specification says none of them has any impact on the calculated score.","In the cvelistV5 snapshot of November 14, 2026, 1,609 of 46,973 published CVE records with a CNA cvssV4_0 entry (3.4%) set at least one Supplemental metric. 149 of 326 CNAs that publish v4.0 scores used them at least once.","Usage is concentrated: Juniper and Palo Alto Networks each account for 258 entries, while VulDB (14,115 v4.0 entries) set none and VulnCheck (11,610) set two. Absence of a Supplemental value tells you about the CNA, not the vulnerability.","The JSON property names are inconsistent (Safety, Automatable and Recovery are capitalised; valueDensity, vulnerabilityResponseEffort and providerUrgency are camelCase), and in 61 CNA entries the vector string and JSON fields disagree on which metrics are set. Parse both and log mismatches.","Treat Automatable and Provider Urgency as the supplier's view at publication time. Palo Alto's own CVSS for CVE-2024-0012 says AU:N while CISA's SSVC entry in the same record says Automatable: yes."],"title":"CVSS v4 Supplemental Metrics in CVE Records: Who Sets Them","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-14","url":"https://firewallsync.com/posts/cvss-v4-supplemental-metrics-in-cve-records-automatable-provider-urgency/"},{"author":"FirewallSync Editorial","description":"MCP security fundamentals: server trust, tool poisoning, credential scope, and the confused deputy problem developers need to address before connecting AI to tools.","faq":[],"markdown_url":"https://firewallsync.com/posts/mcp-security-what-developers-need-to-secure-before-connecting-ai-to-tools/index.md","published":"2026-11-13","reading_minutes":4,"summary":"MCP solved a real integration problem — connecting AI applications to tools and data through a common interface — and in doing so created a new, concentrated attack surface at exactly that connection point.","tags":["ai security"],"takeaways":[],"title":"MCP Security: What Developers Need to Secure Before Connecting AI to Tools","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-13","url":"https://firewallsync.com/posts/mcp-security-what-developers-need-to-secure-before-connecting-ai-to-tools/"},{"author":"Craig Leslie","description":"Notion says Notion AI chat transcripts from under 1% of workspaces were exposed in a Fireworks AI breach. What is confirmed, the timeline, and what to check.","faq":[{"a":"Notion says Notion AI chat transcripts, meaning any model input or output written to a Notion AI chat, were affected for fewer than 1% of workspaces. It says AI Meeting Notes audio, transcripts and summaries were not affected unless used as input in a Notion AI chat, and that record, workspace, user and thread IDs it added for analysis were not exposed.","q":"What data did the Fireworks AI breach expose at Notion?"},{"a":"Fireworks says a third-party cloud provider that hosted its R\u0026D machines had a security incident on October 23, 2026, and that the unauthorized party used that access to reach Fireworks development machines and run an unauthorized workload on October 24. Fireworks has not said how the provider was compromised. A separate earlier incident, identified on October 6, involved unauthorized use of internal Fireworks credentials.","q":"How did attackers get into Fireworks AI?"},{"a":"Notion says workspace owners who were contacted can open Security incident review from the help menu in a desktop browser to download affected-member and, with a verified domain, affected-content CSVs. It recommends treating the data as exposed and immediately rotating any passwords, API keys or other credentials pasted into Notion AI. Notion says customers it has not contacted need take no action now.","q":"What should Notion customers do?"},{"a":"Notion says its zero data retention commitment did not cover data it had to store to provide features such as chat history and Insights, and that during the investigation it learned data was retained from the Fireworks endpoints it used. Neither company has publicly said which Fireworks feature stored the transcripts.","q":"Did Notion's zero data retention promise fail?"}],"markdown_url":"https://firewallsync.com/posts/notion-ai-chats-exposed-fireworks-ai-breach-what-we-know/index.md","published":"2026-11-13","reading_minutes":8,"summary":"Notion says chat transcripts from fewer than 1% of workspaces were exposed after attackers reached its inference vendor Fireworks AI. What Notion and Fireworks have confirmed, the October timeline, and what to check if your product sends prompts to an AI vendor.","tags":["hacking news","notion","fireworks ai","data breach","llmjacking","ai vendors","zero data retention","third-party risk"],"takeaways":["Notion says a security incident at its AI vendor Fireworks AI affected fewer than 1% of Notion workspaces and exposed Notion AI chat transcripts: model inputs and outputs written in Notion AI chats.","Notion says its zero data retention commitment did not cover data it had to store for features such as chat history and Insights, and that its investigation found data was retained at the Fireworks endpoints it used.","Fireworks says a third-party cloud provider hosting its R\u0026D machines had an incident on October 23, and the intruder ran an unauthorized workload on October 24. It links the activity to a financially motivated LLMJacking group it does not name.","Notion says it disconnected from Fireworks as of November 3, deleted data held there, disabled Insights and told customers to treat the data as exposed and rotate any credentials pasted into Notion AI.","For engineering teams the lesson is per-feature retention: ask each AI vendor which endpoints store content (batch jobs, datasets, evaluations), and rotate leaked secrets at the issuer, not just in the vendor's secret store."],"title":"Notion AI Chats Exposed in Fireworks AI Breach: What We Know","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-13","url":"https://firewallsync.com/posts/notion-ai-chats-exposed-fireworks-ai-breach-what-we-know/"},{"author":"John Marshall","description":"CVE JSON metrics: 31.3% of 2026 CVEs carry a CNA cvssV4_0 score, 21.1% no CNA CVSS at all, and baseScore can hide Threat metrics. How to pick one score safely.","faq":[{"a":"No. In CVE_Record_Format.json the CNA published container requires providerMetadata, descriptions, affected and references; metrics is optional. When it is present it is an array with at least one entry, and each entry must contain at least one of cvssV4_0, cvssV3_1, cvssV3_0, cvssV2_0 or other. In 2026, 21.1% of published records had no CNA CVSS entry.","q":"Is the metrics field required in a CVE record?"},{"a":"No. The schema has only baseScore for v4.0, so producers that set Threat metrics such as E:P make a choice. In the 2026 corpus, 5,816 CNA v4.0 entries with Exploit Maturity set stored the Base-only score and 1,067 stored the Base+Threat score. Recompute from vectorString if the distinction matters to your policy.","q":"Does cvssV4_0.baseScore always contain the CVSS-B score?"},{"a":"It describes the situation a metrics entry applies to. The schema says that if no specific scenario is given, GENERAL is the default and applies when no more specific metric matches. In 2026 records, 3,543 CNA metrics entries named a specific scenario, for example F5's Appliance Mode score for CVE-2026-41225.","q":"What does the scenarios field mean in CVE metrics?"},{"a":"Use the CNA score when there is one and keep the CISA-ADP score as a second, labelled value. CISA-ADP mostly fills gaps: it supplied v3.1 for 15,148 of the 20,408 2026 records with no CNA CVSS. Where both existed, 58 of 468 records had different v3.1 base scores, sometimes far apart.","q":"Should I use the CNA score or the CISA-ADP score?"}],"markdown_url":"https://firewallsync.com/posts/cve-json-cvss-metrics-cvssv4-0-cvssv3-1-basescore-trap/index.md","published":"2026-11-13","reading_minutes":10,"summary":"How the metrics array in a CVE JSON 5 record actually looks across all 96,879 CVEs published in 2026: which CVSS versions CNAs ship, when CISA-ADP fills the gap, what scenarios mean, why cvssV4_0.baseScore is sometimes a Base+Threat score, and a tested script that picks one score per record and says where it came from.","tags":["vulnerability management","cve","cve json 5","cvss","cvss v4","cisa vulnrichment","adp","parsing"],"takeaways":["In the CVE JSON 5 schema, metrics is optional in the CNA container. When present it is an array; each entry carries one scoring format (cvssV4_0, cvssV3_1, cvssV3_0, cvssV2_0 or other) plus optional scenarios.","Of 96,879 CVE records published in 2026 (cvelistV5 snapshot of November 13, 2026), 31.3% had a CNA cvssV4_0 entry, 66.4% a CNA cvssV3_1 entry and 21.1% no CNA CVSS at all. CISA-ADP supplied a CVSS v3.1 score for 15,148 of those 20,408 records; 5,260 had no CVSS from either.","The schema's cvssV4_0 object has a baseScore field and no Threat or Environmental score field. Of 6,954 CNA v4.0 entries whose vector sets Exploit Maturity, 5,816 stored the Base-only score and 1,067 stored the Base+Threat (CVSS-BT) score in baseScore.","Pick deterministically: prefer the CNA over ADP, the GENERAL scenario over specific ones, v4.0 over v3.x, and recompute from vectorString when you need to know what baseScore contains. Keep the source with the number.","Where both the CNA and CISA-ADP published v3.1, the scores differed in 58 of 468 records; do not average them, store both."],"title":"CVE JSON CVSS Metrics: cvssV4_0, cvssV3_1, baseScore Trap","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-13","url":"https://firewallsync.com/posts/cve-json-cvss-metrics-cvssv4-0-cvssv3-1-basescore-trap/"},{"author":"FirewallSync Editorial","description":"Excessive agency in AI agents explained: the three sources OWASP identifies — excessive functionality, permissions, and autonomy — and how to actually scope each one down.","faq":[],"markdown_url":"https://firewallsync.com/posts/excessive-agency-in-ai-agents-how-too-much-permission-becomes-a-security-risk/index.md","published":"2026-11-12","reading_minutes":5,"summary":"Excessive agency isn't about an agent doing something malicious on its own. It's about an agent being granted more autonomy, functionality, or permission than its actual task requires, and that surplus becoming the thing an attacker uses.","tags":["ai security"],"takeaways":[],"title":"Excessive Agency in AI Agents: How Too Much Permission Becomes a Security Risk","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-12","url":"https://firewallsync.com/posts/excessive-agency-in-ai-agents-how-too-much-permission-becomes-a-security-risk/"},{"author":"John Marshall","description":"Maersk confirmed data was extracted from a legacy system from a past acquisition on Nov 11, 2026. What is confirmed, what WhiteArmor claims, what to check.","faq":[{"a":"Maersk said on November 11, 2026 that it is investigating an IT security incident involving unauthorized access to, and data extraction from, a local legacy system from a previous acquisition. It says the incident is isolated to that system.","q":"Was Maersk hacked in November 2026?"},{"a":"Maersk says it does not expect the incident to affect operations and that customers can continue to use its platforms to book and manage shipments as usual.","q":"Are Maersk shipping operations affected?"},{"a":"A group calling itself WhiteArmor listed Maersk on its leak site and claims to have stolen data, according to Security Affairs and Cybernews. Maersk has not confirmed any link between the group and the incident.","q":"Who is behind the Maersk attack?"},{"a":"Nothing reported links the two. NotPetya in 2017 was a destructive wiper that shut down Maersk's IT; the 2026 incident, as Maersk describes it, is data extraction from one isolated legacy system.","q":"Is this related to the 2017 NotPetya attack on Maersk?"}],"markdown_url":"https://firewallsync.com/posts/maersk-cyberattack-legacy-system-data-theft-what-we-know/index.md","published":"2026-11-12","reading_minutes":6,"summary":"Maersk says attackers accessed and extracted data from a local legacy system inherited through an acquisition, that operations are unaffected and that its findings so far suggest no sensitive data was disclosed. A group calling itself WhiteArmor claims far more.","tags":["hacking news","maersk","data breach","ransomware","mergers and acquisitions","legacy systems","supply chain"],"takeaways":["On November 11, 2026, Maersk said it is investigating unauthorized access to, and data extraction from, a local legacy system it inherited through a previous acquisition (Reuters, The Register).","Maersk says the incident is isolated to that system, its findings so far strongly suggest no sensitive data was disclosed, and it does not expect any effect on operations.","A group calling itself WhiteArmor claims about 100 GB from the LF Logistics platform and a $5 million ransom demand. Maersk has not confirmed the claims or the link to LF Logistics.","The engineering lesson is inventory: systems that arrive through acquisitions keep their own credentials, integrations and password storage until someone retires or rebuilds them."],"title":"Maersk Cyberattack: Data Stolen From Acquired Legacy System","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-12","url":"https://firewallsync.com/posts/maersk-cyberattack-legacy-system-data-theft-what-we-know/"},{"author":"Craig Leslie","description":"vexctl creates OpenVEX documents, appends statements with add, merges them as a timeline and wraps them in in-toto attestations. Flags and pitfalls in v0.4.4.","faq":[{"a":"vexctl is the OpenVEX project's command-line tool for creating VEX documents, appending statements to them, merging several documents, wrapping a document in a signed in-toto attestation and filtering SARIF scanner results with VEX data.","q":"What is vexctl used for?"},{"a":"No. In v0.4.4 the merge command keeps every statement that matches the optional --product and --vuln filters and sorts them by vulnerability name, then timestamp. The newest statement wins only when a consumer applies the OpenVEX rule of reading the latest statement for a product and vulnerability.","q":"Does vexctl merge pick the newest status?"},{"a":"Use vexctl add with the document path, product, vulnerability and status, for example vexctl add --in-place doc.vex.json pkg:... CVE-... fixed. It appends the statement, increments the document version, sets last_updated and refuses to add a statement dated earlier than existing ones.","q":"How do I add a statement to an existing OpenVEX document?"},{"a":"Attestation subjects need an image reference or a digest. vexctl converts oci purls to image references and keeps products with sha256 or sha512 hashes; other purls without hashes are skipped with a warning, or cause an error if you passed them explicitly as subjects.","q":"Why does vexctl attest skip some products?"}],"markdown_url":"https://firewallsync.com/posts/vexctl-create-add-merge-attest-openvex-documents/index.md","published":"2026-11-12","reading_minutes":8,"summary":"A source-checked guide to vexctl v0.4.4: the create, add, merge and attest commands, the validation rules they enforce, why merge keeps every statement instead of resolving them, and the defaults that quietly produce weak VEX documents.","tags":["vex","openvex","vexctl","sigstore","in-toto","sbom","supply chain security","vulnerability management"],"takeaways":["vexctl v0.4.4 (tagged June 2026) has four commands that matter for producers: create (one statement, new document), add (append a statement, bump version), merge (combine documents) and attest (wrap a document in an in-toto attestation).","Set --author on every command. The default is the literal string \"Unknown Author\", and the OpenVEX spec says author must identify an individual or organization.","For status affected, --action-statement defaults to \"No action statement provided\", which satisfies the validator but tells consumers nothing. Always pass -a with real remediation text.","vexctl merge does not resolve conflicts. It copies every matching statement into one document, sorted by vulnerability name and then timestamp; consumers then take the latest statement per product and vulnerability.","vexctl attest uses predicate type https://openvex.dev/ns/v0.2.0, takes subjects from oci purls or hashes in the document, and --attach always implies --sign."],"title":"vexctl: Create, Add, Merge and Attest OpenVEX Documents","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-12","url":"https://firewallsync.com/posts/vexctl-create-add-merge-attest-openvex-documents/"},{"author":"FirewallSync Editorial","description":"Why giving an AI agent tool access fundamentally changes its threat model, from a text-generation risk to a real-world action risk, and what that requires defensively.","faq":[],"markdown_url":"https://firewallsync.com/posts/ai-agent-security-why-giving-an-agent-tools-changes-the-threat-model/index.md","published":"2026-11-11","reading_minutes":5,"summary":"A chatbot's worst-case failure is generating bad text. An agent's worst-case failure is taking a bad action, with real tool access, based on reasoning that isn't fully predictable in advance.","tags":["ai security"],"takeaways":[],"title":"AI Agent Security: Why Giving an Agent Tools Changes the Threat Model","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-11","url":"https://firewallsync.com/posts/ai-agent-security-why-giving-an-agent-tools-changes-the-threat-model/"},{"author":"Craig Leslie","description":"N-able pulled hosted N-central offline after Able2Shell exploits went public Nov 8. Seven CVEs, a hotfix, no confirmed customer compromise. What to check.","faq":[{"a":"Able2Shell is the name the researcher gave to the exploit chains published on GitHub on November 8, 2026 against N-able N-central. Huntress describes 33 critical vulnerabilities forming four chains: three lead to remote code execution and one to privilege escalation to root. The research targeted N-central Appliance 2026.4.0.31; earlier versions are likely affected, per Huntress.","q":"What is Able2Shell?"},{"a":"Yes. CSO Online reports N-able shipped an emergency hotfix, N-central 2026.4 HF2, on November 10, and that customers must also upgrade Agents and Probes to restore compatibility. N-able's seven CVE records, published the same day, list N-central versions before 26.4.1 as affected. N-able has not said how many of the 33 published issues it reproduced.","q":"Is there a patch for the N-central Able2Shell vulnerabilities?"},{"a":"Not confirmed for customer environments. N-able said it had not seen active exploitation as of November 9 at 15:15 UTC, and TechCrunch quotes it as having 'no evidence of successful exploitation against any of its customer environments.' watchTowr reported exploitation attempts against N-central honeypots after the disclosure. None of the seven new CVEs is in CISA's KEV catalog as of version 2026.11.10.","q":"Has N-central been exploited in the wild?"},{"a":"We could not confirm restoration at the time of writing. N-able's status page said hosted N-central was offline 'while we take the necessary actions to secure the environments and implement appropriate remediation,' as quoted by TechCrunch on November 10. Check N-able's status page for the current state.","q":"Is the hosted N-central service back online?"}],"markdown_url":"https://firewallsync.com/posts/n-able-n-central-able2shell-zero-days-offline-what-we-know/index.md","published":"2026-11-11","reading_minutes":6,"summary":"A researcher published exploit code for what Huntress counts as 33 critical N-able N-central flaws on November 8. N-able took hosted N-central offline, told on-premises customers to shut their servers down, and on November 10 published seven CVEs and a hotfix. What is confirmed, what is not, and what MSPs and IT teams should check.","tags":["hacking news","n-able","n-central","rmm","msp security","zero-day","able2shell","incident response"],"takeaways":["On Sunday, November 8, 2026 an independent researcher published exploit code on GitHub for flaws in N-able N-central, a remote monitoring and management (RMM) platform used by managed service providers. Huntress counts 33 critical vulnerabilities forming four exploit chains ('Able2Shell'): three to remote code execution, one to root.","N-able took its hosted N-central environments offline and told on-premises customers to keep servers offline until it confirmed they could be restored. As of November 9, 15:15 UTC, N-able said it had not seen active exploitation.","On November 10 N-able, as CVE Numbering Authority, published seven CVE records. Two are unauthenticated network flaws scored CVSS v4 9.3: XXE (CVE-2026-125883) and SSRF (CVE-2026-125881). CSO Online reports the fix as N-central 2026.4 HF2; the CVE records list versions before 26.4.1 as affected.","watchTowr says it saw exploitation attempts against N-central honeypots after the disclosure. That is attempted exploitation, not a confirmed customer compromise.","If you run N-central: keep it offline until you can patch, upgrade agents and probes with it, then review logins, new admin accounts, remote sessions and scripts pushed since November 8."],"title":"N-able N-central Offline After Able2Shell Zero-Days","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-11","url":"https://firewallsync.com/posts/n-able-n-central-able2shell-zero-days-offline-what-we-know/"},{"author":"John Marshall","description":"Trivy --vex oci reads OpenVEX attestations from OCI referrers or cosign .att tags but does not verify signatures. How discovery works and how to verify first.","faq":[{"a":"It makes Trivy look in the image's registry for an OpenVEX attestation attached to the scanned image and apply its statements, so findings marked not_affected or fixed are suppressed. Trivy v0.76.2 checks OCI 1.1 referrers first and falls back to the legacy cosign .att tag. Add --show-suppressed to see what VEX filtered.","q":"What does trivy --vex oci do?"},{"a":"No. Trivy's VEX attestation documentation says it 'does not verify the signature of the discovered VEX attestation' and that anyone able to push to the image's repository can attach one that suppresses findings. Use --vex oci only with registries you trust, or verify with cosign first and pass the extracted document as a local file.","q":"Does Trivy verify the signature on a VEX attestation?"},{"a":"Not in Trivy at the moment. Its documentation says a product ID with a tag qualifier, such as pkg:oci/trivy?repository_url=ghcr.io/aquasecurity/trivy\u0026tag=0.50.0, is not supported. Use the repository (all tags) or a specific digest as the purl version.","q":"Can an OpenVEX product ID target an image tag?"},{"a":"In cosign v3.1.4, --type openvex maps to https://openvex.dev/ns. Trivy accepts that exact value or any value that starts with https://openvex.dev/ns/, such as https://openvex.dev/ns/v0.2.0.","q":"What predicate type does cosign use for --type openvex?"}],"markdown_url":"https://firewallsync.com/posts/trivy-vex-oci-openvex-attestations-cosign-attest/index.md","published":"2026-11-11","reading_minutes":9,"summary":"How Trivy's --vex oci finds an OpenVEX document attached to a container image: the three registry layouts it reads, the in-toto.io/predicate-type annotation, why it takes the first match, and the warning that matters most: Trivy does not verify the attestation's signature. Includes a verify-then-scan pattern with cosign verify-attestation.","tags":["vex","openvex","trivy","cosign","sigstore","oci referrers","container security","supply chain security"],"takeaways":["cosign attest --predicate vex.json --type openvex \u003cimage\u003e attaches an OpenVEX document to an image as an in-toto attestation with predicate type https://openvex.dev/ns (cosign v3.1.4 source). Since cosign v3 it is stored as an OCI 1.1 referrer by default.","trivy image --vex oci \u003cimage\u003e looks for OCI 1.1 referrers first (Sigstore bundle, DSSE envelope or bare in-toto statement), then falls back to the legacy cosign v2 sha256-\u003cdigest\u003e.att tag. Referrer support arrived in Trivy v0.73.0; the feature is still marked experimental.","Trivy's documentation warns that it does not verify the signature of a discovered VEX attestation: anyone who can push to the repository can attach one that suppresses findings.","In Trivy v0.76.2 the first referrer that decodes to OpenVEX wins, in registry order; other VEX attestations on the same image are not merged. Keep one current VEX document per image digest.","For anything that gates a release, verify with cosign verify-attestation --type openvex and a pinned identity, extract the predicate, and pass the file to trivy --vex \u003cfile\u003e instead of --vex oci."],"title":"Trivy --vex oci: OpenVEX Attestations on OCI Images","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-11-11","url":"https://firewallsync.com/posts/trivy-vex-oci-openvex-attestations-cosign-attest/"},{"author":"FirewallSync Editorial","description":"How LLM applications leak sensitive data in practice — training data exposure, cross-user context bleed, and overshared retrieval — and how to actually prevent it.","faq":[],"markdown_url":"https://firewallsync.com/posts/llm-data-leakage-how-ai-applications-accidentally-expose-sensitive-information/index.md","published":"2026-11-10","reading_minutes":5,"summary":"Most LLM data leakage isn't caused by a model being hacked. It's caused by ordinary application design decisions that hand the model, or its output, more than the current user should see.","tags":["ai security"],"takeaways":[],"title":"LLM Data Leakage: How AI Applications Accidentally Expose Sensitive Information","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-10","url":"https://firewallsync.com/posts/llm-data-leakage-how-ai-applications-accidentally-expose-sensitive-information/"},{"author":"John Marshall","description":"Vanderbilt ordered devices shut down on Nov 9 after a cyberattack claimed by the BYOD ransomware gang, its third US university in a month. Confirmed vs claimed.","faq":[{"a":"Vanderbilt told students and employees to shut down university devices and disconnect from campus Wi-Fi while its IT team worked to contain a cybersecurity incident, according to FOX 17 and Gizmodo. Its 7 p.m. update said Wi-Fi was available again but 'access to most Vanderbilt systems and applications remains restricted', and that classes would run on a normal schedule on November 10.","q":"What happened at Vanderbilt University on November 9, 2026?"},{"a":"A ransomware group calling itself BYOD claimed responsibility online. Vanderbilt's public updates have not named an attacker, and the claim has not been independently verified.","q":"Who is behind the Vanderbilt cyberattack?"},{"a":"Not confirmed. BYOD claims it obtained 'Over 1 Million SSNs, Full Names, Email addresses, Phone Numbers, Student \u0026 Employee Records', as quoted by Gizmodo. FOX 17 reports that Vanderbilt has not confirmed whether sensitive information was stolen.","q":"Was student data stolen in the Vanderbilt attack?"},{"a":"According to FOX 17, Vanderbilt Health, which operates separately from the university, said its IT systems were unaffected and patient care continued. It told employees to stay off university IT systems until further notice.","q":"Is Vanderbilt Health affected?"}],"markdown_url":"https://firewallsync.com/posts/vanderbilt-university-byod-ransomware-attack-what-we-know/index.md","published":"2026-11-10","reading_minutes":6,"summary":"Vanderbilt University told students and staff on November 9 to shut down their devices and get off campus Wi-Fi while it contained a cybersecurity incident. A new ransomware group called BYOD claims it stole over a million SSNs, and it has hit Michigan State and UNM Health in the last month. What is confirmed, what is only claimed, and what engineers can take from the three incidents.","tags":["hacking news","ransomware","byod ransomware","higher education","incident response","vanderbilt"],"takeaways":["On November 9, 2026 Vanderbilt University told students and employees to shut down university devices and disconnect from campus Wi-Fi while it contained a cybersecurity incident. By that evening Wi-Fi was back but most systems and applications were still restricted.","A ransomware group calling itself BYOD claims responsibility and says it took 'Over 1 Million SSNs' plus student and employee records. Vanderbilt has not confirmed any data theft, and the claim is unverified.","BYOD, which The Record says emerged in September, has also claimed Michigan State University (1.3 million people, per the group) and UNM Health, where ambulances were diverted. In both cases the institutions' own statements describe far less than the group claims.","Lessons for engineers: have a tested 'shut it all down' message and an out-of-band help channel ready, watch synced storage such as OneDrive and Box for mass renames, and don't size your response to an attacker's leak-site claims."],"title":"Vanderbilt BYOD Ransomware Attack: What We Know So Far","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-10","url":"https://firewallsync.com/posts/vanderbilt-university-byod-ransomware-attack-what-we-know/"},{"author":"Craig Leslie","description":"KEV knownRansomwareCampaignUse flips from Unknown to Known silently: 114 times since Jan 2025, none reversed. Lag data, vendors and a script to catch flips.","faq":[{"a":"It is a string field on every KEV entry. CISA's JSON schema says it is 'Known' if the vulnerability is known to have been leveraged as part of a ransomware campaign, and 'Unknown' if CISA lacks confirmation that it has been used for ransomware. Unknown means unconfirmed, not confirmed absent.","q":"What does knownRansomwareCampaignUse mean in the CISA KEV catalog?"},{"a":"Not in a way you can rely on. The change shows up as an edited value in the catalog files. In the kev-data history from January 2025 to November 2026 the field changed 114 times, and dateAdded and dueDate stay the same, so a flip is only visible if you compare the field between catalog versions.","q":"Does CISA announce when a KEV entry changes to Known ransomware use?"},{"a":"For the 50 entries that were both added and flipped between January 27, 2025 and November 10, 2026, the median was 20 days after dateAdded, 17 flipped within 7 days and 39 within 90 days. Across all 114 flips, including entries added before 2025, the median was 352 days.","q":"How long after a KEV listing does the ransomware flag usually flip?"},{"a":"In catalog version 2026.11.09, 361 of 1,773 entries (20.4%) have knownRansomwareCampaignUse set to Known. Among entries added in 2026, 27 are Known, and 25 of those were listed as Unknown first.","q":"What share of KEV entries are marked Known for ransomware?"}],"markdown_url":"https://firewallsync.com/posts/kev-knownransomwarecampaignuse-unknown-to-known-flips/index.md","published":"2026-11-10","reading_minutes":7,"summary":"CISA's KEV catalog marks each entry's ransomware use as Known or Unknown, and it changes that field without an announcement. We replayed every kev-data commit since January 2025: 114 entries flipped from Unknown to Known, none flipped back, and 25 of the 27 Known entries added in 2026 started as Unknown. Here is how often it happens, how long it takes, and a tested script to catch the next flip.","tags":["cisa kev","knownransomwarecampaignuse","ransomware","vulnerability prioritization","vulnerability management","security operations"],"takeaways":["knownRansomwareCampaignUse has two values. CISA's schema defines 'Known' as known to have been leveraged in a ransomware campaign and 'Unknown' as CISA lacking confirmation, so Unknown is not 'no'.","Replaying 341 kev-data commits (January 27, 2025 to November 10, 2026), 114 entries changed from Unknown to Known: 59 in 2025 and 55 so far in 2026. No entry changed from Known back to Unknown.","New entries are almost always listed as Unknown: 521 of the 524 entries that first appeared in the window. 50 of those 521 later flipped, with a median of 20 days after dateAdded; 17 flipped within a week.","Flips also hit old entries: the median gap across all 114 is 352 days, and two Win32k CVEs flipped 1,771 days after they were added. Microsoft accounts for 33 of the 114.","A flip does not change dateAdded or dueDate, so a pipeline that only looks for new cveIDs never sees it. Diff the field between catalog versions and route Unknown-to-Known changes into your ransomware-driven SLA."],"title":"KEV knownRansomwareCampaignUse: 114 Unknown-to-Known Flips","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-10","url":"https://firewallsync.com/posts/kev-knownransomwarecampaignuse-unknown-to-known-flips/"},{"author":"FirewallSync Editorial","description":"Indirect prompt injection explained: how attackers plant instructions in documents, emails, and web pages your AI application will later process on someone else's behalf.","faq":[],"markdown_url":"https://firewallsync.com/posts/indirect-prompt-injection-the-ai-security-problem-hiding-in-your-data/index.md","published":"2026-11-09","reading_minutes":5,"summary":"The attacker doesn't need to talk to your AI application at all. They just need to get malicious text into something it will eventually read — a document, a webpage, an inbox.","tags":["ai security"],"takeaways":[],"title":"Indirect Prompt Injection: The AI Security Problem Hiding in Your Data","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-09","url":"https://firewallsync.com/posts/indirect-prompt-injection-the-ai-security-problem-hiding-in-your-data/"},{"author":"Craig Leslie","description":"Supermarket freezer hacks: attackers set refrigeration controllers to heat in France, then Belgium and Sweden. What CISA and the FBI advise operators to do.","faq":[{"a":"Attackers remotely took control of store refrigeration controllers and changed their settings so units heated or defrosted continuously. The initial access method has not been confirmed publicly. The Record reports that the October attack in France compromised internet-exposed remote controllers, and the US joint warning says actors may exploit unpatched or internet-accessible controllers through authentication bypass, code injection or weak configurations.","q":"How were supermarket freezers hacked in Europe?"},{"a":"France in the first wave around October 11-12, 2026. The second wave on November 2 was reported in Belgium, Sweden, Luxembourg, Estonia, the Netherlands and Ireland. Later reports mention incidents in Germany, Italy, Portugal and Romania, but we have not confirmed those against primary sources.","q":"Which countries were affected by the refrigeration cyberattacks?"},{"a":"No source has linked it. CVE-2025-41450 (authentication bypass) and CVE-2025-41451 (command injection) affect Danfoss AK-SM 8xxA controllers before versions 4.2 and 4.3.1. Neither is in CISA's KEV catalog as of version 2026.11.06, and CISA's enrichment of both records still says Exploitation: poc. If you run these controllers, patch them anyway.","q":"Is CVE-2025-41450 being exploited in the supermarket attacks?"},{"a":"Nobody has claimed them and no government has attributed them. Belgium's Centre for Cybersecurity said it is not known who is behind the attack and that it may never be possible to identify them. It is also not known whether the same actor carried out both waves.","q":"Who is behind the supermarket freezer attacks?"}],"markdown_url":"https://firewallsync.com/posts/supermarket-freezer-hacks-refrigeration-controllers-cisa-fbi-warning/index.md","published":"2026-11-09","reading_minutes":7,"summary":"Attackers remotely switched supermarket freezers to heating in hundreds of French stores in October and in Belgium, Sweden, Estonia and other countries on November 2. On November 5 CISA, the FBI and three other US agencies told operators to pull refrigeration controllers off the internet. What is confirmed, what is not, and what to check.","tags":["hacking news","ot security","ics","refrigeration controllers","cisa","danfoss"],"takeaways":["In a first wave around October 11-12, 2026, attackers took control of refrigeration controllers in French supermarkets and pushed units into heating and defrost; the Paris prosecutor's office opened an investigation and, per The Record, counts around 700 victims.","A second wave on November 2 hit stores in Belgium, Sweden, Luxembourg, Estonia and the Netherlands. Retailers including Delhaize, Carrefour Belgium, ICA and Coop confirmed affected stores; Swedish stores reported freezers set to 50°C.","On November 5, CISA, the FBI, HHS, USDA and DoT published a joint warning that attackers may use 'unpatched or internet-accessible controllers' to force 'continuous heating and defrost cycles'. It names no vendor and no CVE.","Danfoss equipment was reported in France and Estonia, Copeland equipment in Belgium. Both vendors said after the French attack that affected devices were mainly discontinued products. How attackers got in has not been confirmed publicly.","What to do: take controller panels off the public internet, change default passwords, retire end-of-life controllers, alert on setpoint changes, and keep a temperature alarm that does not depend on the controller."],"title":"Supermarket Freezer Hacks: What the CISA-FBI Warning Says","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-09","url":"https://firewallsync.com/posts/supermarket-freezer-hacks-refrigeration-controllers-cisa-fbi-warning/"},{"author":"John Marshall","description":"KEV time to listing: for CVEs added in 2026, CISA's dateAdded trails the record's datePublished by a median 14 days, 4 for same-year CVEs. Data and script.","faq":[{"a":"Measured from the CVE record's datePublished to KEV's dateAdded, the median for entries added in 2026 is 14 days (catalog 2026.11.06). For 2026 CVE IDs added in 2026 it is 4 days, and 28% were listed on or before the day the record was published. CISA adds entries when it has reliable evidence of exploitation, so the gap measures when evidence arrived, not how slow CISA is.","q":"How long does CISA take to add a vulnerability to KEV?"},{"a":"Yes. 62 KEV entries have a dateAdded earlier than their record's datePublished, including CVE-2026-118527 (Ivanti Connect Secure), added on October 30, 2026 with a record published on November 3. A pipeline that only checks KEV for CVEs it has already ingested from the CVE List or NVD will miss these.","q":"Can a CVE be in KEV before the CVE record is published?"},{"a":"Yes, regularly. Of the 285 entries added in 2026, 51 were for CVEs published more than 365 days earlier, the oldest being CVE-2008-4128 (Cisco), about 17.8 years after publication. New exploitation evidence for an old bug is a reason to recheck systems where that CVE was accepted as low risk.","q":"Does CISA add old CVEs to KEV?"},{"a":"Both, for different questions. The KEV dueDate is set from dateAdded, so compliance tracking starts there. Exposure starts at publication, and for 2026 CVEs the median wait for a KEV listing was 4 days, so waiting for KEV before patching internet-facing systems gives up that head start.","q":"Should my patch SLA start at CVE publication or at KEV listing?"}],"markdown_url":"https://firewallsync.com/posts/kev-time-to-listing-cve-published-to-dateadded-days-by-vendor/index.md","published":"2026-11-09","reading_minutes":9,"summary":"How long after a CVE record is published does CISA add it to KEV? We joined all 1,769 KEV entries to their CVE records' datePublished. For 2026 additions the median gap is 14 days, but the catalog mixes same-week listings with backfills of decade-old CVEs. Numbers by year and vendor, a tested script, and what it means for SLA clocks.","tags":["cisa kev","cve record","vulnerability prioritization","patch sla","vulnerability management","security operations"],"takeaways":["Joining KEV catalog 2026.11.06 (1,769 entries) to each CVE record's cveMetadata.datePublished: for the 285 entries added in 2026, the median gap from publication to KEV dateAdded is 14 days.","That median hides two populations. The 197 2026 additions with 2026 CVE IDs had a median gap of 4 days, and 60% were listed within 7 days. Meanwhile 51 of the 285 (18%) were CVEs published more than a year earlier.","Vendor matters: for 2025-2026 additions, Microsoft's median gap was 1 day (49% listed on or before publication day), while Linux kernel entries had a median of 132 days and none within a week.","62 KEV entries, 4 of them in 2026, were listed before their CVE record was published. Match KEV by cveID, never by whether you already have a CVE or NVD record.","For SLAs: a KEV due date starts at dateAdded, but your exposure started at publication. Track both clocks, and treat a KEV listing of an old CVE as a trigger to reopen accepted risk."],"title":"KEV Time to Listing: CVE datePublished to dateAdded","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-09","url":"https://firewallsync.com/posts/kev-time-to-listing-cve-published-to-dateadded-days-by-vendor/"},{"author":"FirewallSync Editorial","description":"Practical defenses against prompt injection in LLM applications: input handling, output validation, and — most importantly — limiting what a successful attack can actually do.","faq":[],"markdown_url":"https://firewallsync.com/posts/how-to-defend-llm-applications-against-prompt-injection/index.md","published":"2026-11-08","reading_minutes":5,"summary":"There's no single fix that closes prompt injection completely. What actually reduces risk is layering several partial mitigations and, critically, limiting what a successful injection could accomplish.","tags":["ai security"],"takeaways":[],"title":"How to Defend LLM Applications Against Prompt Injection","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-08","url":"https://firewallsync.com/posts/how-to-defend-llm-applications-against-prompt-injection/"},{"author":"Craig Leslie","description":"DigiCert says self-registered accounts on an internet-exposed internal system read archived customer data and some ID files in 2026. CA systems unaffected.","faq":[{"a":"According to DigiCert, as reported by BleepingComputer and Cybersecurity News, no. The company says the unauthorized account could not issue certificates, approve validations or modify records, and that issued certificates and customer login credentials remain secure.","q":"Do DigiCert customers need to replace their TLS certificates?"},{"a":"DigiCert says the affected archived records included names, business email addresses, phone numbers and business addresses. For some customers who had supplied them for validation, identity documents such as passports, driver's licences or utility bills were also involved. DigiCert said it was still reviewing which individuals' documents were affected and would notify them.","q":"What data was exposed in the DigiCert breach?"},{"a":"Not publicly known. The reports we reviewed do not name an actor, and a DigiCert spokesperson declined to tell The Record whether the attackers had contacted the company. DigiCert said it had reported the incident to law enforcement.","q":"Who was behind the DigiCert breach?"},{"a":"DigiCert's account, as reported, places the start of the exposure at a February 2025 migration and the fix in early July 2026, about 17 months. The unauthorized data access itself is reported for June 27 to July 2, 2026.","q":"How long was the DigiCert system exposed?"}],"markdown_url":"https://firewallsync.com/posts/digicert-data-breach-self-registration-archived-customer-data/index.md","published":"2026-11-08","reading_minutes":6,"summary":"DigiCert says an internal business system, moved out of a VPN-only network in February 2025, became reachable from the internet. Self-registered accounts then read archived customer records, including some identity documents, between June 27 and July 2, 2026. Certificates and CA systems were not affected, according to the company. What is confirmed, what is not, and what engineers should check in their own estate.","tags":["hacking news","data breach","digicert","certificate authority","access control","authorization"],"takeaways":["DigiCert says a February 2025 migration moved an internal business system out of a VPN-restricted network and left it reachable from the internet.","According to DigiCert's incident report, as reported by BleepingComputer and The Record, accounts created through a self-registration feature had read-only access to archived customer information because of an authorization flaw; that data was accessed between June 27 and July 2, 2026.","Exposed data: names, business email addresses, phone numbers and business addresses, plus identity documents (passports, driver's licences, utility bills) for some customers. DigiCert has not said how many people are affected.","DigiCert says its CA services, CertCentral and DigiCert ONE were not compromised, and that issued certificates and customer login credentials remain secure.","The engineering lesson is not about certificates: a network move removed the only control in front of an app whose self-registration and permission checks were never designed for the internet."],"title":"DigiCert Data Breach: Self-Registered Accounts Read ID Files","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-11-08","url":"https://firewallsync.com/posts/digicert-data-breach-self-registration-archived-customer-data/"},{"author":"John Marshall","description":"cvelistV5's deltaLog.json lists every new and updated CVE record for about 30 days. Read fetchTime, new and updated, dedupe by cveId, re-baseline when stale.","faq":[{"a":"delta.json holds only the latest change batch written by the repository's update job. deltaLog.json is an array of those batches, newest first, retained for a rolling 30 days according to the repository README and the update workflow's CVES_DEFAULT_DELTA_LOG_HISTORY_IN_DAYS setting. On November 8, 2026, delta.json was identical to the first element of deltaLog.json.","q":"What is the difference between delta.json and deltaLog.json in cvelistV5?"},{"a":"No. The repository tooling puts a CVE in the new queue when its JSON file is added to the repository, and in updated when an existing file is modified. A reserved ID that is rejected without ever being published still arrives as a new file. Check cveMetadata.state in the record itself.","q":"Does 'new' in cvelistV5 deltas mean a newly published CVE?"},{"a":"The update workflow is scheduled every five minutes, but scheduled runs do not land that regularly. Across the 30 days of deltaLog we read, the median gap between batches was about 10 minutes, the 90th percentile about 29 minutes and the longest gap about 3 hours 49 minutes.","q":"How often does cvelistV5 update?"},{"a":"The log can no longer tell you everything that changed. Download the most recent midnight baseline zip from the repository's releases (or git pull a clone), replace your mirror, and restart the checkpoint from the newest fetchTime in deltaLog.json.","q":"What if my sync has been down for more than 30 days?"}],"markdown_url":"https://firewallsync.com/posts/cvelistv5-delta-json-deltalog-incremental-cve-sync/index.md","published":"2026-11-08","reading_minutes":9,"summary":"The CVE Program's cvelistV5 repository publishes two small index files, cves/delta.json and cves/deltaLog.json, that tell you which CVE records changed and when. We read 30 days of deltaLog (2,870 entries, 122,761 changes) to show what each field means, what 'new' and 'updated' really cover, how often the feed moves, and a tested Python script that keeps a local mirror current from a checkpoint.","tags":["cve","cvelistv5","cve json 5","vulnerability management","data feeds","automation"],"takeaways":["cves/delta.json is the most recent change batch only; cves/deltaLog.json is a newest-first array of those batches covering a rolling 30 days. Each batch has fetchTime, numberOfChanges, new, updated and error.","Each change item carries cveId, cveOrgLink, githubLink (the raw JSON URL of the record) and dateUpdated. It does not say what changed: fetch the record.","'new' means a record file that did not exist in the repository before, not 'a newly disclosed vulnerability': 28 of 300 sampled IDs from the new queue are REJECTED records today.","Most 'updated' entries are enrichment, not CNA edits: in 244 of 299 sampled updated records, the record's dateUpdated matches the CISA-ADP container's timestamp.","Sync from a stored fetchTime checkpoint, keep the newest dateUpdated per cveId, and fall back to the midnight baseline zip when the checkpoint is older than the log's oldest entry."],"title":"cvelistV5 delta.json and deltaLog.json: Sync CVE Changes","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-08","url":"https://firewallsync.com/posts/cvelistv5-delta-json-deltalog-incremental-cve-sync/"},{"author":"FirewallSync Editorial","description":"Prompt injection explained: why LLMs can't reliably separate instructions from data, the difference between direct and indirect attacks, and why it's structurally hard to fully fix.","faq":[],"markdown_url":"https://firewallsync.com/posts/prompt-injection-explained-how-attackers-manipulate-ai-applications/index.md","published":"2026-11-07","reading_minutes":5,"summary":"Prompt injection has topped OWASP's LLM Top 10 for two consecutive editions, for a structural reason: instructions and data flow through the same channel, and the model has no reliable way to tell them apart.","tags":["ai security"],"takeaways":[],"title":"Prompt Injection Explained: How Attackers Manipulate AI Applications","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-11-07","url":"https://firewallsync.com/posts/prompt-injection-explained-how-attackers-manipulate-ai-applications/"},{"author":"Craig Leslie","description":"Linux kernel CVE records split affected data into git ranges and X.Y.* stable-branch ranges. Read both: 1,056 of 7,987 2026 records need the git entry.","faq":[{"a":"For a kernel.org kernel, take the version (for example 6.12.114), and evaluate it against both product entries in the CVE record's affected block, using the CVE schema rule that the first matching entry wins and unlisted versions take defaultStatus. If either entry says affected, treat it as affected. For a distribution kernel (anything after a dash, such as 6.12.114-1-amd64), use your distribution's security tracker instead.","q":"How do I check if my Linux kernel is affected by a CVE?"},{"a":"It appears on the last version entry of nearly every 2026 kernel record, as a range from the mainline release that contains the fix up to '*', marked unaffected. Neither the CVE schema's examples nor the kernel's CVE process document defines it, so a strict parser will not know how to compare it. Treating it as a plain kernel version comparison is an assumption you should make explicitly.","q":"What does versionType original_commit_for_fix mean in a kernel CVE record?"},{"a":"One common cause is matching on the record's cpeApplicability block. In the 2026 kernel records, those CPE ranges are OR'd and all start at the release that introduced the bug, so the widest range covers every fixed stable release below the mainline fix. In our check, 97.4% of the fixed stable releases named in the records matched as vulnerable that way.","q":"Why does a scanner flag my patched stable kernel?"},{"a":"The kernel's CVE process document says that, because almost any kernel bug might be exploitable, the assignment team is 'overly cautious and assign CVE numbers to any bugfix that they identify'. It also says applicability is up to the user, and that most systems use only a small part of the source tree.","q":"Why are there so many Linux kernel CVEs?"}],"markdown_url":"https://firewallsync.com/posts/linux-kernel-cve-records-affected-versions-stable-branches/index.md","published":"2026-11-07","reading_minutes":11,"summary":"Every Linux kernel CVE record carries two product entries: git commit ranges and kernel version ranges. We read all 7,987 kernel CNA records published in 2026 to show how the two fit together, where reading only one goes wrong (1,056 records), why the CPE block over-matches fixed kernels, and a tested Python checker for a kernel.org version.","tags":["linux kernel","cve","cve json 5","vulnerability management","affected versions","kernel cna"],"takeaways":["All 7,987 Linux kernel CNA records published in 2026 (to November 6) have exactly two product entries. In 7,923 of them the first holds git commit ranges with defaultStatus unaffected and the second holds kernel version ranges with defaultStatus affected.","The version entry reads top to bottom and the first match wins: the release that introduced the bug, everything before it unaffected, one 'X.Y.N to X.Y.*' unaffected range per fixed stable branch, then the mainline fix release onward. Anything not listed falls back to affected.","In 1,056 records the git entry also holds kernel version ranges marking a stable release as affected where the version entry alone says unaffected, usually because the bug was backported into that stable branch. Evaluate both entries.","The records' OR'd cpeApplicability ranges matched 35,340 of the 36,296 fixed stable releases they name (97.4%) as vulnerable. Use the affected block, not CPE, for kernel.org versions.","None of this applies directly to distribution kernels: kernel.org says anything after the dash in a version means a distribution kernel. Use the distribution's own tracker or VEX data for those."],"title":"Linux Kernel CVE Records: Read Stable-Branch Affected Ranges","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-07","url":"https://firewallsync.com/posts/linux-kernel-cve-records-affected-versions-stable-branches/"},{"author":"John Marshall","description":"EPSS for KEV CVEs starts low: 49% of entries added in the last 90 days sit below 0.1, and late risers took a median 14 days to cross it. Audit script inside.","faq":[{"a":"EPSS estimates the probability of exploitation activity in the next 30 days from automated data, while KEV is a human-curated list of CVEs with reliable evidence of exploitation. In our data the gap is largest for new listings: on the day CISA added an entry, the median EPSS score was 0.0076, and half of the entries added in the last 90 days are still below 0.1.","q":"Why do some CISA KEV vulnerabilities have a low EPSS score?"},{"a":"None that is useful. On November 6, 2026, 54 KEV CVEs scored below 0.01 and one KEV entry had no EPSS score at all. A cutoff low enough to catch every KEV entry would flag a large share of all CVEs. Check KEV membership separately, then use EPSS for the CVEs without exploitation evidence.","q":"What EPSS threshold catches all KEV vulnerabilities?"},{"a":"In our sample it usually did, with a lag. For 113 KEV entries added between June 16 and October 7, 2026, the share at or above 0.1 rose from 16.5% on the day of listing to 33.6% after 7 days and 60.2% after 30 days. We cannot tell from the score files which inputs moved each score, so this shows timing, not cause.","q":"Does EPSS go up after CISA adds a CVE to KEV?"},{"a":"Download the KEV JSON from cisagov/kev-data and a daily EPSS file from empiricalsec/epss_scores, then run the audit script in this article with your cutoff. It reports how many KEV CVEs fall below the cutoff, split by how recently they were listed.","q":"How do I check my own EPSS threshold against KEV?"}],"markdown_url":"https://firewallsync.com/posts/epss-scores-for-kev-cves-low-after-listing-threshold-audit/index.md","published":"2026-11-07","reading_minutes":7,"summary":"An EPSS cutoff of 0.1 misses a quarter of CISA's KEV catalog today, and half of the entries added in the last 90 days. We tracked 113 KEV additions through 145 daily EPSS v5 files: the median score on the day of listing was 0.0076, and entries that crossed 0.1 after listing took a median of 14 days. Data, a tested audit script and the rule it implies.","tags":["epss","cisa kev","vulnerability prioritization","vulnerability management","security operations"],"takeaways":["Joining CISA KEV catalog 2026.11.06 to the EPSS file of November 6, 2026: 436 of 1,768 scored KEV CVEs (24.7%) sit below 0.1, and 54 (3.1%) below 0.01.","Recency is the driver. Of 106 KEV entries added in the last 90 days, 52 (49.1%) are below 0.1, against 22.5% for entries listed more than a year ago.","For 113 KEV entries added June 16 to October 7, 2026, the median EPSS on the day of listing was 0.0076. Only 16.5% of those scored that day were at or above 0.1; 30 days later it was 60.2%.","Of the 79 entries that reached 0.1 by November 6, 15 were already there the day before listing; the other 64 took a median of 14 days. 34 never reached it.","The rule this supports: KEV membership must route a CVE on its own. A pipeline that folds KEV into an EPSS cutoff is slowest exactly when an exploited CVE is newest."],"title":"EPSS Scores for KEV CVEs: Why New Entries Score Low","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-07","url":"https://firewallsync.com/posts/epss-scores-for-kev-cves-low-after-listing-threshold-audit/"},{"author":"FirewallSync Editorial","description":"Workload identity explained: how it replaces static cloud keys with short-lived, automatically issued credentials, and how to actually migrate to it.","faq":[],"markdown_url":"https://firewallsync.com/posts/what-is-workload-identity-and-why-should-you-stop-using-static-cloud-keys/index.md","published":"2026-11-06","reading_minutes":5,"summary":"A static cloud key stored on a server is a credential that works until someone remembers to rotate it. Workload identity replaces that with a credential that's issued fresh, per request, and expires by design.","tags":["iam","cloud security"],"takeaways":[],"title":"What Is Workload Identity and Why Should You Stop Using Static Cloud Keys?","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-11-06","url":"https://firewallsync.com/posts/what-is-workload-identity-and-why-should-you-stop-using-static-cloud-keys/"},{"author":"John Marshall","description":"EPSS v5 went live on June 15, 2026. CVEs at or above 0.1 fell 26% overnight and top-1,000 overlap was 588. Diff data, threshold fixes and a model_version guard.","faq":[{"a":"The daily EPSS file carried model_version v2026.06.15 for the first time on June 15, 2026. The June 14 file was the last scored by v2025.03.14 (v4). Empirical Security announced v5 on May 13, 2026, and published beta files labelled v2026.05.12 between May 22 and June 8.","q":"When did EPSS v5 start?"},{"a":"If the alert uses a fixed probability cutoff, v5 put far fewer CVEs above it. In our comparison, the count of CVEs at or above 0.1 fell 26.1% and at or above 0.5 fell 40.9% between the last v4 file and the first v5 file. Nothing about those vulnerabilities changed that day; the model did.","q":"Why did my EPSS alerts drop after June 15, 2026?"},{"a":"Percentile keeps a stable count (the top 10% is always about 10% of scored CVEs), so it survives model changes better for queue sizing. Probability is the calibrated output and the one to use for expected-value reasoning, but any fixed cutoff needs re-baselining whenever model_version changes.","q":"Should I use EPSS percentile or probability after the v5 change?"},{"a":"Read the first line of each daily CSV. Since v2 it is a comment of the form #model_version:...,score_date:.... Compare model_version to the value from your previous run, and if it differs, pause threshold-driven alerts, diff the two files and re-baseline before resuming.","q":"How do I detect the next EPSS model change?"}],"markdown_url":"https://firewallsync.com/posts/epss-v5-score-shift-model-version-v2026-06-15/index.md","published":"2026-11-06","reading_minutes":8,"summary":"EPSS v5 (model v2026.06.15) replaced v4 in the daily feed on June 15, 2026. We diffed the last v4 file against the first v5 file: 26% fewer CVEs at or above 0.1, 41% fewer at or above 0.5, a median score almost three times higher, and only 588 of the top 1,000 CVEs unchanged. Here is what that means for threshold rules, and a tested guard that stops your pipeline on the next model change.","tags":["epss","epss v5","vulnerability management","vulnerability prioritization","security operations"],"takeaways":["The EPSS daily file switched from model v2025.03.14 (v4) to v2026.06.15 (v5) on June 15, 2026. The first line of every file names the model, so a pipeline can detect the change without any outside notice.","Comparing the June 14 (v4) and June 15 (v5) files: 23,069 CVEs scored 0.1 or higher before the switch and 17,041 after, a 26.1% drop. At 0.5 the count fell from 7,193 to 4,249 (40.9%).","Low scores moved up and high scores moved down. The median rose from 0.00262 to 0.00751, while the 99th percentile fell from 0.772 to 0.595. 71.5% of CVEs scored on both days moved by a factor of two or more.","Percentile rules kept their size by design, but not their membership: 9,837 CVEs crossed into the 90th percentile and 9,944 dropped out of it in one day, against 25 and 29 the day before.","KEV-listed CVEs mostly kept their flags: of 1,619 entries added by June 14, 1,224 scored 0.1 or higher under v4 and 1,219 under v5. Their median score still fell from 0.704 to 0.535."],"title":"EPSS v5: How Scores Shifted and What to Re-baseline","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-06","url":"https://firewallsync.com/posts/epss-v5-score-shift-model-version-v2026-06-15/"},{"author":"Craig Leslie","description":"OpenVEX products and subcomponents decide which findings a VEX statement suppresses. How go-vex and Trivy match purls, versions, qualifiers and timestamps.","faq":[{"a":"The OpenVEX spec defines subcomponents as any components possibly included in the product where the vulnerability originates. They are usually the product's dependencies, they should carry software identifiers such as purls, and they should also appear in the product's SBOM.","q":"What is a subcomponent in OpenVEX?"},{"a":"In go-vex v0.2.9 and Trivy v0.76.2, yes. If the purl in the VEX document has no version, it matches any version of the same type, namespace and name. If it has a version, the scanned purl must have exactly that version; there is no version-range support.","q":"Does an OpenVEX purl without a version match all versions?"},{"a":"The most common causes are a purl that differs in type, namespace or name; a qualifier in the VEX purl (such as arch or repository_url) that the scanned purl lacks or has with another value; a version in the VEX purl that differs from the installed one; or a newer matching statement whose status is affected or under_investigation.","q":"Why does Trivy ignore my OpenVEX statement?"},{"a":"go-vex sorts matching statements by timestamp, using the document timestamp for statements that have none, and Trivy applies the last one. That is why a new statement added to an existing document should keep older statements' original timestamps, as the OpenVEX spec's update example shows.","q":"Which statement wins when an OpenVEX document has several for the same CVE?"}],"markdown_url":"https://firewallsync.com/posts/openvex-products-subcomponents-purl-matching-trivy/index.md","published":"2026-11-06","reading_minutes":9,"summary":"An OpenVEX statement suppresses a finding only when its products, subcomponents and purls line up with what the scanner found. Here is how the go-vex library behind Trivy v0.76.2 matches them: versionless purls, qualifier subsets, subcomponent scoping, timestamp inheritance and which status wins, with a schema-valid example document and a tested Python port of the rules.","tags":["openvex","vex","purl","sbom","trivy","supply chain","vulnerability management"],"takeaways":["In OpenVEX v0.2.0 the product is the software the statement is about and subcomponents are where the vulnerability originates, usually its dependencies. Both are matched by @id, by the identifiers map (purl, cpe22, cpe23) or by hashes.","go-vex v0.2.9, which Trivy v0.76.2 uses, treats the purl in the VEX document as the general pattern: no version matches every version, and every qualifier it lists must appear in the scanned purl with the same value. Extra qualifiers on the scanned side are ignored.","A product with no subcomponents matches regardless of which package carries the CVE. Listing subcomponents narrows the statement to findings in those packages. That is the difference between suppressing a CVE across a whole image and suppressing it in one package.","When several statements match, statements are sorted by timestamp, a statement without one inherits the document timestamp, and the newest wins. Trivy suppresses only when that winner is not_affected or fixed.","Trivy walks from the vulnerable package up through every parent in its dependency graph. A statement on an intermediate module only suppresses the finding if every path to the root passes through a not-affected node."],"title":"OpenVEX Products and Subcomponents: purl Matching Rules","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-06","url":"https://firewallsync.com/posts/openvex-products-subcomponents-purl-matching-trivy/"},{"author":"FirewallSync Editorial","description":"Non-human identities — service accounts, API keys, AI agents — now outnumber human users in most environments, but get a fraction of the identity governance.","faq":[],"markdown_url":"https://firewallsync.com/posts/non-human-identities-the-security-problem-behind-service-accounts-and-ai-agents/index.md","published":"2026-11-05","reading_minutes":5,"summary":"Most identity security programs were designed around human users logging in. Service accounts, API keys, and now AI agents don't fit that model, and it shows in how little oversight they typically get.","tags":["iam","cloud security"],"takeaways":[],"title":"Non-Human Identities: The Security Problem Behind Service Accounts and AI Agents","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-11-05","url":"https://firewallsync.com/posts/non-human-identities-the-security-problem-behind-service-accounts-and-ai-agents/"},{"author":"Craig Leslie","description":"CVE JSON 5.2 packageURL: a version-less purl in affected[]. 6,284 of 400,445 records use it. How to normalize it and join it to SBOM purls, with tested Python.","faq":[{"a":"It is an optional field on each object in the affected array of a CVE JSON 5.2 record. It holds a Package URL (purl) such as pkg:npm/fast-uri that identifies the affected package. The schema forbids a version inside it; the affected versions are stated in the same entry's versions array.","q":"What is packageURL in a CVE record?"},{"a":"CVE Record Format 5.2.0. The proposal is RFD 0002, Supporting Package URLs in CVE, in the CVEProject/cve-schema repository, and the field is present in the 5.2.0 schema. Records that declare dataVersion 5.1 cannot carry it.","q":"Which CVE schema version added packageURL?"},{"a":"You can find candidate records that way, but you still have to evaluate the versions array and defaultStatus to know whether your version is affected. The CVE purl has no version, so a purl match only says the record is about your package.","q":"Can I match a CVE to my SBOM using only the purl?"},{"a":"Most often because of case or normalization differences (pkg:pypi/GitPython against pkg:pypi/gitpython, or uppercase GitHub owners), because your SBOM purl has qualifiers like arch or distro, or because the CNA used a different purl type, such as pkg:github instead of the package registry type.","q":"Why does my purl not match the purl in a CVE record?"}],"markdown_url":"https://firewallsync.com/posts/cve-packageurl-purl-cve-json-5-2-affected-sbom-matching/index.md","published":"2026-11-05","reading_minutes":9,"summary":"CVE Record Format 5.2.0 added an optional packageURL field to affected entries. We counted it across all 400,445 published records: 1.6% carry one, 12 CNAs produce them, and a third of GitHub purls are not in canonical lowercase. Here is what the field means, how to normalize it, and a tested Python join against a CycloneDX SBOM.","tags":["cve","cve json 5","purl","package url","sbom","vulnerability management","version matching","appsec"],"takeaways":["packageURL is an optional field on each affected[] product in CVE Record Format 5.2.0. The schema says it \"MUST NOT include a version\"; versions stay in the versions array.","It does not replace the identifier pair. Every affected entry still needs vendor and product, or collectionURL and packageName. packageURL sits alongside them.","In the cvelistV5 snapshot of 5 November 2026, 6,284 of 400,445 published records (1.6%) had a packageURL in the CNA container, from 12 CNAs. VulnCheck produced 78.8% of them. No ADP container carried one.","Adoption is rising fast: 2,000 of the 17,058 records first published in October 2026 (11.7%) carry one. None of the 7,324 purls we found embedded a version.","Normalize before you join. 756 of 2,371 pkg:github purls (31.9%) used uppercase letters the purl spec says should be lowercased, and GitPython appears as both pkg:pypi/GitPython and pkg:pypi/gitpython across records."],"title":"CVE packageURL: Join CVE JSON 5.2 Records to SBOM purls","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-05","url":"https://firewallsync.com/posts/cve-packageurl-purl-cve-json-5-2-affected-sbom-matching/"},{"author":"John Marshall","description":"CSAF VEX product_status lists hold product IDs, not purls. Walk product_tree branches and relationships to purls, then join to your SBOM. Tested Python inside.","faq":[{"a":"CSAF 2.0 is the OASIS standard for machine-readable security advisories. CSAF VEX is one of its profiles: a CSAF document whose document.category is csaf_vex and which carries a product tree, vulnerabilities with notes and a CVE or other ID, and at least one fixed, known_affected, known_not_affected or under_investigation status, plus impact and action statements where the profile requires them.","q":"What is the difference between CSAF and CSAF VEX?"},{"a":"In the product tree. Each full product name, whether it sits at the end of a branch, in full_product_names or inside a relationship, can carry product_identification_helper.purl. The vulnerability's product_status lists only reference product IDs, so you have to resolve each ID back to its product tree entry.","q":"Where is the purl in a CSAF VEX file?"},{"a":"No. In Trivy v0.76.1 the CSAF matcher builds its status map from known_not_affected and fixed only. A match on either removes the finding; known_affected and under_investigation statements leave the scanner result as it was.","q":"Does Trivy use known_affected from a CSAF VEX document?"},{"a":"Usually because a status was attached to a bare component ID instead of a relationship ID, or because the consumer matched only the component purl and ignored the parent product. A not_affected statement on a component alone reads as true for that component in every product, which is rarely what the vendor meant.","q":"Why does a CSAF VEX statement match the wrong products?"}],"markdown_url":"https://firewallsync.com/posts/csaf-vex-product-tree-product-status-purl-sbom-matching/index.md","published":"2026-11-05","reading_minutes":13,"summary":"A CSAF VEX file never puts a purl next to a status. Product IDs in product_status point into the product_tree, often through relationships. Here is how to walk branches and relationships, collapse the eight status lists into four VEX answers, and join the result to a CycloneDX SBOM, with a tested Python script.","tags":["csaf","vex","csaf vex","sbom","purl","vulnerability management","trivy","security operations"],"takeaways":["In CSAF 2.0, product_status holds product IDs. The purl that lets you match your SBOM lives in the product_tree, under branches, full_product_names or relationships, in product_identification_helper.","A relationship such as default_component_of creates a new product ID that means \"this component inside that product\". A status on that ID should only apply when both the component and the parent product match your inventory.","CSAF defines eight product_status lists. For a VEX decision, collapse them to four: affected (first_affected, known_affected, last_affected), not affected, fixed (first_fixed, fixed) and under investigation. Treat recommended as a pointer, not a status.","The VEX profile requires an impact statement (a flag or an impact threat) for every known_not_affected product and an action statement in remediations for every known_affected product. Missing ones mean the file does not satisfy the profile.","Trivy v0.76.1 applies only known_not_affected and fixed from CSAF, matches by purl, and treats default_component_of, installed_on and installed_with the same way. known_affected and under_investigation entries never change its output."],"title":"CSAF VEX product_tree: Resolve Product IDs to SBOM purls","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-05","url":"https://firewallsync.com/posts/csaf-vex-product-tree-product-status-purl-sbom-matching/"},{"author":"FirewallSync Editorial","description":"How to audit unused cloud accounts, roles, and service identities systematically, using activity data rather than manual review, before they become someone else's access.","faq":[],"markdown_url":"https://firewallsync.com/posts/how-to-audit-unused-cloud-accounts-roles-and-service-identities/index.md","published":"2026-11-04","reading_minutes":4,"summary":"An unused identity with live credentials is functionally identical to a vulnerability sitting unpatched — nobody's using it, but it still works exactly as well for whoever finds it.","tags":["iam","cloud security"],"takeaways":[],"title":"How to Audit Unused Cloud Accounts, Roles, and Service Identities","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-11-04","url":"https://firewallsync.com/posts/how-to-audit-unused-cloud-accounts-roles-and-service-identities/"},{"author":"John Marshall","description":"CVE-2026-119977 lets attackers hijack WordPress 4.7-7.1.3 core updates. Block /wp-content/upgrade/, then update to 7.1.4 or your branch fix. KEV due 6 Nov.","faq":[{"a":"WordPress 7.1.4, released 28 October 2026, and backports on older branches including 7.0.8, 6.9.11, 6.8.12, 6.7.11, 6.6.11, 6.5.14 and down to 4.7.39. WordPress 4.6 and earlier no longer receive security updates. WordPress says only the most recent version is actively supported.","q":"Which WordPress versions fix CVE-2026-119977?"},{"a":"CISA lists it in the Known Exploited Vulnerabilities catalog (added 3 November 2026), and CISA's SSVC assessment records Exploitation as active. Wordfence told The Hacker News it blocked 62,969 attacks aimed at the flaw in 24 hours, while noting that not every request used the exact vulnerable path.","q":"Is CVE-2026-119977 exploited in the wild?"},{"a":"Minor core auto-updates are enabled by default in WordPress unless a site turns them off with the WP_AUTO_UPDATE_CORE constant, a filter or the auto_update_core_minor option, so many sites will already have moved to their branch's fix. Check the version on every site rather than assuming.","q":"Does WordPress install the fix automatically?"},{"a":"Deny web access to the /wp-content/upgrade/ path in the web server configuration, so files staged there during an update cannot be requested, and avoid running core updates until you can. Put the rule in server config or the site's root .htaccess, not inside wp-content/upgrade, because affected versions empty that folder before each update.","q":"What can I do if I cannot update yet?"}],"markdown_url":"https://firewallsync.com/posts/wordpress-cve-2026-119977-upgrade-race-rce-kev/index.md","published":"2026-11-04","reading_minutes":7,"summary":"CISA added CVE-2026-119977 to the KEV catalog on 3 November 2026. In WordPress 4.7 through 7.1.3, a full-package core update stages a copy of WordPress, installer included, in a web-served folder under wp-content/upgrade, and an unauthenticated attacker who reaches it during the update window can run code. Here are the fixed versions per branch, what the 7.1.4 code change does, why the update itself is a risky moment, and what to check afterwards.","tags":["wordpress","cve-2026-119977","kev","rce","race condition","php","patch management"],"takeaways":["CVE-2026-119977 affects WordPress core 4.7.0 through 7.1.3. The CNA (Wordfence) describes predictable, web-accessible staging paths used during full-package core upgrades that let unauthenticated attackers execute code during the upgrade window. CVSS 3.1 is 8.1 (AC:H).","Fixed releases shipped on 28 October 2026: 7.1.4, plus backports for every branch from 4.7 (4.7.39) to 7.0 (7.0.8). In all 25 branches the fix is the release after the last affected version in the CVE record.","CISA added it to KEV on 3 November 2026 with a federal due date of 6 November and forensic triage required. CISA's SSVC entry is Exploitation: active, Automatable: no, Technical Impact: total.","The 7.1.4 change randomises the staging folder name with 16 characters, keeps an index.php in wp-content/upgrade so the folder cannot be listed, and refuses to copy a wp-config.php from the staged package into the live site.","The update to the fixed release is itself a core update run by the old, vulnerable staging code. Block web requests to /wp-content/upgrade/ at the web server before you update, then check for unexpected files."],"title":"WordPress CVE-2026-119977: Patch the Upgrade Race RCE","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-04","url":"https://firewallsync.com/posts/wordpress-cve-2026-119977-upgrade-race-rce-kev/"},{"author":"Craig Leslie","description":"cpeApplicability in a CVE JSON record: how nodes, AND/OR operators, negate, vulnerable and versionStart/End bounds combine, with a tested Python evaluator.","faq":[{"a":"It is an optional array in a CVE JSON 5 record, added in CVE Record Format 5.1.1, where the CNA (or an ADP) describes affected products with the CPE Applicability Language: CPE 2.3 match strings grouped into nodes joined by AND or OR, each marked vulnerable or not, with optional version bounds.","q":"What is cpeApplicability in a CVE record?"},{"a":"affected[].cpes is a flat list of CPE names attached to one product entry. cpeApplicability is a separate block that can combine CPEs with AND/OR logic, mark platform CPEs as not vulnerable, negate nodes and express version ranges with versionStartIncluding, versionStartExcluding, versionEndIncluding and versionEndExcluding. The schema's cpes description itself recommends the newer block.","q":"What is the difference between affected.cpes and cpeApplicability?"},{"a":"No. It uses nearly the same structure, but it is written by the CNA inside the CVE record. NVD configurations are NVD's own analysis published through the NVD API. The two can differ, so record which source a match came from.","q":"Is cpeApplicability the same as the NVD configuration?"},{"a":"Not by default. The schema recommends that the block align with the affected block as much as possible, but CPE version ranges cannot say \"fixed in this stable branch\". Evaluate both and send disagreements to a human or the vendor advisory.","q":"Should a scanner trust cpeApplicability over the affected versions list?"}],"markdown_url":"https://firewallsync.com/posts/cve-cpeapplicability-cpe-match-nodes-operators-version-ranges/index.md","published":"2026-11-04","reading_minutes":10,"summary":"Since CVE Record Format 5.1.1, a CNA can put CPE Applicability Language statements in a cpeApplicability block. Here is how nodes, operator, negate, vulnerable and the four version bounds combine, what a 2,000-record sample of 2026 CVEs contains, and a tested Python evaluator that shows where the block and the affected block disagree.","tags":["vulnerability management","cve","cve json 5","cpe","cpeApplicability","version matching","appsec"],"takeaways":["cpeApplicability is an optional array in the CNA container (and allowed in ADP containers) of a CVE JSON record. The schema says it is \"mostly copied/forked from the NIST NVD CVE API v2.0 schema\", so it reads like an NVD configuration.","Each element holds nodes; each node has an operator (AND or OR, required), an optional negate flag and a cpeMatch list. Each cpeMatch needs criteria (a CPE 2.3 name) and vulnerable, plus up to four versionStart/End bounds.","vulnerable: false marks a platform or context CPE, such as the operating system a vulnerable application must run on. A match that only hits non-vulnerable entries does not mean the host is affected.","In a random sample of 2,000 records from 2026 (1,969 published), 27.3% carried a CNA cpeApplicability block; four CNAs (Linux, VulnCheck, Microsoft, Oracle) produced 88.7% of them, and no ADP container carried one.","CPE ranges cannot express stable-branch fixes. In a Linux kernel record we tested, cpeApplicability marks 6.18.9 vulnerable while the same record's affected block marks 6.18.8 to 6.18.* unaffected. Prefer the affected block and use CPE as a cross-check."],"title":"CVE cpeApplicability: CPE Match Nodes, Operators, Ranges","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-04","url":"https://firewallsync.com/posts/cve-cpeapplicability-cpe-match-nodes-operators-version-ranges/"},{"author":"FirewallSync Editorial","description":"Why identity has become the primary security boundary in cloud environments, and where teams still over-invest in network controls at the expense of IAM.","faq":[],"markdown_url":"https://firewallsync.com/posts/cloud-iam-vs-network-security-why-identity-has-become-the-new-perimeter/index.md","published":"2026-11-03","reading_minutes":4,"summary":"Network segmentation still matters, but in a cloud environment, the thing standing between an attacker and your data is increasingly a permission check, not a firewall rule.","tags":["iam","cloud security"],"takeaways":[],"title":"Cloud IAM vs Network Security: Why Identity Has Become the New Perimeter","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-11-03","url":"https://firewallsync.com/posts/cloud-iam-vs-network-security-why-identity-has-become-the-new-perimeter/"},{"author":"Craig Leslie","description":"CycloneDX VEX mapped to OpenVEX and CSAF: analysis.state, justification, response and BOM-Link affects, with a tested Python converter and the gaps that lose data.","faq":[{"a":"In the vulnerabilities array. Each vulnerability has an id, an affects list whose ref points at a component (by bom-ref in the same BOM or by BOM-Link to another BOM), and an analysis object with state, justification, response and detail.","q":"Where is VEX data in a CycloneDX document?"},{"a":"There is none. CycloneDX defines false_positive as a vulnerability that was falsely identified or associated with the component. OpenVEX and CSAF have only not_affected, affected, fixed and under_investigation. Trivy maps false_positive to not_affected; a converter should do the same and keep the original state in the impact statement.","q":"What is the OpenVEX equivalent of CycloneDX false_positive?"},{"a":"Close enough to map, in our reading. CycloneDX defines code_not_reachable as vulnerable code that is not invoked at runtime; OpenVEX's vulnerable_code_not_in_execute_path covers a product that includes the vulnerable code but does not call or use it.","q":"Does code_not_reachable mean the same as vulnerable_code_not_in_execute_path?"},{"a":"Yes, as an experimental feature. Trivy's documentation says it supports only a separate VEX BOM alongside a CycloneDX SBOM, that BOM-Links must be placed in affects.ref, and that a not_affected state stops the vulnerability being reported.","q":"Can Trivy read CycloneDX VEX?"}],"markdown_url":"https://firewallsync.com/posts/cyclonedx-vex-analysis-state-justification-openvex-csaf-mapping/index.md","published":"2026-11-03","reading_minutes":11,"summary":"CycloneDX VEX uses six analysis states and nine justifications; OpenVEX and CSAF use four statuses and five justifications. Here is a field-by-field mapping, which pairs line up and which do not, how Trivy reads CycloneDX VEX, and a tested Python converter that refuses to invent justifications.","tags":["vex","cyclonedx","openvex","csaf","sbom","vulnerability management","appsec"],"takeaways":["CycloneDX VEX lives in a BOM's vulnerabilities array: analysis.state (6 values), analysis.justification (9 values), analysis.response (5 values) and affects[].ref pointing at a component, usually by BOM-Link.","States map cleanly onto OpenVEX and CSAF statuses except two: false_positive has no counterpart and resolved_with_pedigree collapses into fixed. Trivy's code treats false_positive as not_affected.","Only two of CycloneDX's nine justifications line up with the five CISA justifications used by OpenVEX and CSAF: code_not_present and code_not_reachable. The rest have no clean equivalent.","OpenVEX lets a not_affected statement carry an impact_statement instead of a justification label, so a converter can keep the CycloneDX reason as text rather than guess a label.","A CycloneDX VEX that points at components by BOM-Link only applies to the exact SBOM serial number and version it names. Trivy skips statements whose BOM-Link does not match the SBOM it scans."],"title":"CycloneDX VEX: analysis.state and justification to OpenVEX","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-03","url":"https://firewallsync.com/posts/cyclonedx-vex-analysis-state-justification-openvex-csaf-mapping/"},{"author":"John Marshall","description":"CVE ADP containers explained: what CISA-ADP, the CVE Program Container and supplier ADPs (redhat-SADP, siemens-SADP) add to a CVE record, and how to merge them.","faq":[{"a":"It is an object in containers.adp holding information about the CVE added by an Authorized Data Publisher rather than by the CNA that assigned the ID. The CVE JSON schema says multiple ADPs can provide information for one CVE ID, and each container must identify its source in providerMetadata.","q":"What is an ADP container in a CVE record?"},{"a":"No. The CISA Vulnrichment README says CISA will at no point overwrite the originating CNA's data in the CNA container, and that it populates CVSS or CWE in its own ADP container for CVEs where the CNA has not supplied them. In our sample, 381 of 389 CISA-ADP CVSS entries were on records whose CNA container had no CVSS.","q":"Does CISA-ADP overwrite the CNA's CVSS score?"},{"a":"An ADP-format container with shortName CVE and title CVE Program Container. The cvelistV5 README says it holds references added by the CVE Program after July 31, 2024, while CNA-provided references stay in the CNA container. Records with no Program-added references have no such container.","q":"What is the CVE Program Container?"},{"a":"Supplier ADP containers. In the records we sampled they carry x_adpType supplier and add the supplier's own affected products and versions to a CVE that another CNA assigned, for example a Linux kernel CVE. Siemens describes SADP as a way for a supplier to add information to a vulnerability in an upstream dependency.","q":"What are redhat-SADP and siemens-SADP?"}],"markdown_url":"https://firewallsync.com/posts/cve-adp-containers-cisa-adp-cve-program-supplier-sadp/index.md","published":"2026-11-03","reading_minutes":11,"summary":"A CVE record has one CNA container and possibly several ADP containers. Here is what CISA-ADP, the CVE Program Container and supplier ADPs such as redhat-SADP and siemens-SADP add, how often each appears in a sample of 2,000 2026 records, and a tested script that shows who said what.","tags":["vulnerability management","cve","cve json 5","cisa vulnrichment","adp","ssvc"],"takeaways":["A CVE JSON 5 record has exactly one cna container and an optional adp array. Each ADP container names its publisher in providerMetadata.shortName, and the schema requires only providerMetadata, so an ADP block can be very small.","In a random sample of 2,000 cvelistV5 records for 2026 (1,963 published), 93.9% carried a CISA-ADP container, 2.2% a CVE Program Container and 1.2% a supplier ADP (redhat-SADP or siemens-SADP).","CISA-ADP always carried SSVC in the sample, and added CVSS to 19.8% and a CWE to 17.1% of published records, almost always where the CNA had not supplied one. KEV data appeared in 5 records.","The cvelistV5 README says the CNA container plus the CVE Program Container are the minimum you must read; every other ADP container is optional from the Program's point of view.","Never let an ADP value silently replace a CNA value. Store each value with its provider, apply a documented precedence, and treat supplier ADP affected data as scoped to that supplier's products."],"title":"CVE ADP Containers: CISA-ADP, CVE Program and Supplier ADPs","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-03","url":"https://firewallsync.com/posts/cve-adp-containers-cisa-adp-cve-program-supplier-sadp/"},{"author":"FirewallSync Editorial","description":"How GitHub Actions OIDC works technically, how to configure the AWS trust relationship, and why short-lived tokens change the actual risk profile of CI/CD credentials.","faq":[],"markdown_url":"https://firewallsync.com/posts/github-actions-oidc-stop-storing-cloud-credentials-in-cicd/index.md","published":"2026-11-02","reading_minutes":5,"summary":"A static cloud credential stored as a CI secret is valid until someone notices it's wrong. An OIDC-issued credential is valid for one workflow run and expires whether anyone notices or not.","tags":["appsec","iam"],"takeaways":[],"title":"GitHub Actions OIDC: Stop Storing Cloud Credentials in CI/CD","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-02","url":"https://firewallsync.com/posts/github-actions-oidc-stop-storing-cloud-credentials-in-cicd/"},{"author":"Craig Leslie","description":"Read affected versions in a CVE JSON 5 record: lessThan vs lessThanOrEqual, changes, versionType and the defaultStatus fallback, with a tested Python evaluator.","faq":[{"a":"It is the status for every version of that product that no entry in the versions list matches. The CVE JSON schema says that if defaultStatus is not specified, it defaults to unknown. A record that lists only the affected versions and omits defaultStatus therefore says nothing about other versions.","q":"What does defaultStatus mean in a CVE record?"},{"a":"lessThan is the non-inclusive upper limit, which the schema describes as the least version not in the range; it is usually the fixed version. lessThanOrEqual is the inclusive upper limit, the greatest version in the range. Only one of the two should be used on an entry.","q":"What is the difference between lessThan and lessThanOrEqual?"},{"a":"Not safely. The CVE schema documentation says custom means an otherwise unspecified ordering and that using it means the version status algorithm cannot be executed. Treat a custom range as needing the vendor's own advisory or a vendor-specific comparator.","q":"Can I compare versions with versionType custom automatically?"},{"a":"No. The versions and defaultStatus fields sit in the CNA's affected product entries. CPE data can appear in a product's cpes array or in a separate cpeApplicability block, and NVD adds its own configurations. In our sample, 29.5% of published records carried a cpeApplicability block.","q":"Is this the same as the CPE configuration on NVD?"}],"markdown_url":"https://firewallsync.com/posts/cve-affected-versions-lessthan-defaultstatus-versiontype/index.md","published":"2026-11-02","reading_minutes":10,"summary":"The affected block in a CVE record is a small algorithm, not a list. Here is how version, lessThan, lessThanOrEqual, changes, versionType and defaultStatus combine, what a sample of 1,000 2026 CVE records actually contains, and a tested Python evaluator that refuses version types it cannot order.","tags":["vulnerability management","cve","cve json 5","version matching","sbom","appsec"],"takeaways":["In a CVE JSON 5 record, each product in containers.cna.affected has a versions list and an optional defaultStatus. Together they define a lookup: the first matching entry decides the status, and any version no entry matches gets defaultStatus.","If defaultStatus is missing, the schema says it defaults to unknown. In our random sample of 1,000 CVE records from 2026, 54.8% of 2,061 product entries had no defaultStatus, so any version the CNA did not list is unknown, not safe.","A range entry needs versionType, because \"less than\" depends on the numbering scheme. In the sample, 35.5% of 2,831 ranges used custom, which the schema says means the status algorithm cannot be executed.","lessThan names the first fixed version (exclusive); lessThanOrEqual names the last affected one (inclusive). A * in an upper limit means \"arbitrarily large\", so 6.6.157 to 6.6.* covers the rest of the 6.6 branch.","Code that matches scanner or SBOM versions against CVE records should return three answers plus \"cannot decide\", and send the last two to a human or to the vendor advisory."],"title":"CVE Affected Versions: lessThan, defaultStatus, versionType","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-02","url":"https://firewallsync.com/posts/cve-affected-versions-lessthan-defaultstatus-versiontype/"},{"author":"John Marshall","description":"Vulnerability prioritization workflow from CVE record to patch ticket: record state, affected check, VEX, KEV, CISA SSVC, EPSS, plus a tested Python script.","faq":[{"a":"Validate the CVE record and your affected status first, apply supplier VEX, then check exploitation evidence (KEV and CISA's SSVC Exploitation value), then EPSS for likelihood, and finally combine exploitation and automatability with your own exposure and impact inputs. CVSS describes severity; it is useful context in the ticket but should not decide order on its own.","q":"What order should I check KEV, EPSS, CVSS and SSVC in?"},{"a":"For US federal civilian agencies, KEV entries carry due dates under CISA's binding directives. For everyone else it is strong exploitation evidence, not a deadline. In CERT/CC's deployer table, active exploitation never yields Defer, but it yields Scheduled in six of the 72 rows, all with low or medium Human Impact.","q":"Does a KEV listing always mean patch immediately?"},{"a":"At minimum: asset, product and version, CVE ID and record state, affected status and how it was decided, any VEX status and justification, KEV dateAdded and dueDate, CISA SSVC values with their timestamp, EPSS score, percentile and score date, your exposure and impact inputs, the resulting priority, and the date the data was pulled.","q":"What should a vulnerability ticket contain?"},{"a":"CISA says Vulnrichment covers new and recent CVEs it assesses, so many records have no CISA-ADP SSVC block. Treat that as missing data, not low risk: decide Exploitation yourself from KEV, the vendor advisory and threat reporting, and route the finding to manual triage.","q":"What if a CVE has no CISA SSVC data?"}],"markdown_url":"https://firewallsync.com/posts/vulnerability-prioritization-workflow-cve-record-to-patch-ticket/index.md","published":"2026-11-02","reading_minutes":11,"summary":"A six-step patch-decision flow that takes a scanner finding from CVE record to ticket: check the record state, confirm you are affected, apply supplier VEX, read KEV and CISA's SSVC values, add EPSS, then let your exposure decide. Includes a tested Python script that writes the ticket fields.","tags":["vulnerability management","kev","epss","ssvc","vex","cve","patch management"],"takeaways":["Prioritize in a fixed order of questions: is the CVE record valid, do we run an affected version, has the supplier said we are not affected, is it exploited (KEV, CISA SSVC), how likely is exploitation (EPSS), and how exposed and important is this asset.","The first three steps close tickets rather than rank them. In our six-finding test run, two of six rows closed before any scoring: a rejected CVE ID and a supplier VEX not_affected statement.","Public data supplies only part of the decision. CISA's ADP container gives Exploitation and Automatable; System Exposure and Human Impact are yours to set per asset. With those four values, CERT/CC's deployer table returns Defer, Scheduled, Out-of-cycle or Immediate.","Signals disagree, and the ticket should keep each one. FortiMail CVE-2026-104286 is in KEV with an Immediate deployer result on an internet-facing host, yet its EPSS score in the November 1, 2026 file was 0.04573.","Store the raw fields (state, KEV dates, SSVC values and timestamp, EPSS and model date, VEX status, your exposure inputs) so a reviewer can see why a ticket got its priority, and re-run the flow when any feed changes."],"title":"Vulnerability Prioritization Workflow: CVE Record to Ticket","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-02","url":"https://firewallsync.com/posts/vulnerability-prioritization-workflow-cve-record-to-patch-ticket/"},{"author":"FirewallSync Editorial","description":"How to secure GitHub Actions: workflow permissions, third-party action risk, secrets scoping, and using OIDC instead of long-lived cloud credentials.","faq":[],"markdown_url":"https://firewallsync.com/posts/how-to-secure-github-actions-permissions-secrets-and-oidc/index.md","published":"2026-11-01","reading_minutes":4,"summary":"A GitHub Actions workflow often runs with more access than the task actually requires, simply because the defaults are permissive and nobody scoped them down before shipping.","tags":["appsec","supply chain"],"takeaways":[],"title":"How to Secure GitHub Actions: Permissions, Secrets, and OIDC","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-01","url":"https://firewallsync.com/posts/how-to-secure-github-actions-permissions-secrets-and-oidc/"},{"author":"Craig Leslie","description":"Escape user text before it enters a regex or LIKE pattern: re.escape, RegExp.escape, regexp.QuoteMeta, and LIKE with an ESCAPE clause. Tested code inside.","faq":[{"a":"Use your language's built-in escaper and build the pattern from its output: re.escape in Python, RegExp.escape in JavaScript (Node.js 24 and current browsers), and regexp.QuoteMeta in Go. Do this whenever the user's text should match literally. If users are meant to type real patterns, escaping is the wrong tool; use a linear-time engine and limits instead.","q":"How do I escape user input for a regular expression?"},{"a":"No. A bind parameter makes the value data rather than SQL, so it prevents SQL injection, but LIKE still interprets % and _ inside that value. In our tests a bound value of % matched every row in both SQLite 3.53.1 and PostgreSQL 16.15. Escape the escape character, % and _ in the value, then add an ESCAPE clause.","q":"Do parameterized queries escape LIKE wildcards?"},{"a":"It is available in Node.js 24 (V8 13.6), where we tested it, and MDN marks it Baseline 2025 for current browsers. It is not available in Node.js 22, where typeof RegExp.escape is 'undefined'. Check for it at runtime and use a fallback on older runtimes, keeping the fallback's limits in mind.","q":"Is RegExp.escape available in Node.js?"},{"a":"Escaping makes user text match literally, so the user can no longer add quantifiers or groups to your pattern. It does not fix a vulnerable pattern you wrote yourself, and it is not an option when the product is meant to accept user-written patterns. For that case, see our guide to running user-supplied regex safely.","q":"Does escaping user input prevent ReDoS?"}],"markdown_url":"https://firewallsync.com/posts/escape-user-input-regex-sql-like/index.md","published":"2026-11-01","reading_minutes":7,"summary":"A search box that drops user text into new RegExp() or a LIKE pattern is a small injection bug: a dot matches anything, .* matches everything, and a % turns a filtered search into a full table scan that returns every row. Parameterised queries do not fix it. This guide shows the correct escaping function in Python, JavaScript and Go, the LIKE ESCAPE rules for PostgreSQL and SQLite, and the replacement-string trap that catches people after they have escaped the pattern.","tags":["regex","input validation","sql","like","javascript","python","go","appsec"],"takeaways":["If user text is meant to be matched literally, escape it before it becomes part of a pattern. In our tests an unescaped search for al.ce also matched alice, and .* matched every line.","Use the platform function: Python re.escape, JavaScript RegExp.escape (Node.js 24 has it; Node.js 22 does not), Go regexp.QuoteMeta.","The common hand-rolled JavaScript escaper misses cases. Placed after a backreference or inside a character class, it changed what the pattern matched in our Node.js 24 test; RegExp.escape did not.","Bind parameters stop SQL injection but not LIKE wildcards. A bound value of % still matched every row in SQLite and PostgreSQL. Escape \\, % and _ in the value and state ESCAPE explicitly.","Escaping the pattern is half the job. User text used as a replacement string is interpreted too: $\u0026 in JavaScript and backslash sequences in Python's re.sub. Pass a function instead."],"title":"Escape User Input in Regex and SQL LIKE: Tested Fixes","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-11-01","url":"https://firewallsync.com/posts/escape-user-input-regex-sql-like/"},{"author":"John Marshall","description":"CVE-2026-41709 lets an ESX admin act unlogged. Patch to ESXi80U3j-25429389, 9.0.2.0100 or 9.1.0.0 builds and forward host logs off-box. KEV due 11 Nov.","faq":[{"a":"An insufficient-logging vulnerability (CWE-778) in VMware ESX, published by Broadcom on 29 July 2026 in VMSA-2026-0006. A user who already has administrator access can perform certain operations without them being logged. Broadcom rates it Low, CVSS 3.1 base score 2.7, and has not published which operations are affected.","q":"What is CVE-2026-41709?"},{"a":"ESX 8.0: ESXi80U3j-25429389. ESX 9.0.x: ESXi-9.0.2.0100-25595025. ESX 9.1.x: ESXi-9.1.0.0-25370933. VMware Cloud Foundation 5.x: ESX 5.2.4. Telco Cloud Platform 5.0.x and 5.1.x: follow Broadcom KB449886. The CVE record lists no affected ESX 7.0 versions for this CVE.","q":"Which ESXi build fixes CVE-2026-41709?"},{"a":"KEV lists vulnerabilities with evidence of exploitation, not high scores. CVSS rates the flaw in isolation, and requiring admin privileges pulls the score down. In a real intrusion the attacker often has those privileges already, and a way to act without leaving log entries is valuable to them. CISA's SSVC entry for this CVE records exploitation as active.","q":"Why is a CVSS 2.7 vulnerability in the CISA KEV catalog?"},{"a":"No. The patch stops the gap from being used after it is installed. Anything done through it before then was never logged. Review other evidence for that period, such as vCenter events, remote syslog copies of other ESX logs and changes on the VMs themselves.","q":"Does patching CVE-2026-41709 recover missing logs?"}],"markdown_url":"https://firewallsync.com/posts/vmware-esx-cve-2026-41709-insufficient-logging-kev/index.md","published":"2026-11-01","reading_minutes":7,"summary":"CISA added CVE-2026-41709, a 2.7-rated VMware ESX insufficient-logging flaw, to the KEV catalog on 28 October 2026, three months after Broadcom fixed it in VMSA-2026-0006. A bug that lets an admin act without leaving a log entry matters most after someone else has become that admin. This guide covers the fixed builds, why a Low CVSS score should not decide the priority, and how to make sure ESX logs leave the host.","tags":["vmware","esxi","cve-2026-41709","kev","logging","vsphere","patch management"],"takeaways":["CVE-2026-41709 is an insufficient-logging flaw in VMware ESX: per Broadcom, 'a malicious administrator could exploit this issue to perform certain operations without them being logged.' Broadcom does not say which operations.","Fixed builds per VMSA-2026-0006: ESX 8.0 ESXi80U3j-25429389; 9.0.x ESXi-9.0.2.0100-25595025; 9.1.x ESXi-9.1.0.0-25370933; VCF 5.x ESX 5.2.4; Telco Cloud Platform 5.0.x/5.1.x per KB449886. There is no workaround.","CVSS is 2.7 (Low) because exploitation needs high privileges. CISA's KEV listing (added 28 Oct 2026, due 11 Nov 2026) and its SSVC 'Exploitation: active' value say it is being used anyway, most plausibly to hide activity after an admin account is compromised.","Patching closes the gap going forward; it does not restore entries that were never written. Forward ESX logs to a remote syslog collector so evidence of other actions leaves the host.","The same advisory fixed CVE-2026-59310 in vCenter, which CISA added to KEV in August. If you patched vCenter then, check that your ESX hosts were updated too."],"title":"VMware ESX CVE-2026-41709: Patch the Logging Gap Now","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-11-01","url":"https://firewallsync.com/posts/vmware-esx-cve-2026-41709-insufficient-logging-kev/"},{"author":"FirewallSync Editorial","description":"How dependency confusion attacks work, why they compromised Apple, Microsoft, and PayPal without a single CVE, and how to actually close the gap.","faq":[],"markdown_url":"https://firewallsync.com/posts/dependency-confusion-how-package-managers-can-become-an-attack-path/index.md","published":"2026-10-31","reading_minutes":4,"summary":"Dependency confusion doesn't exploit a vulnerability in any package. It exploits a resolution assumption most package managers share, and it was proven against 35+ major companies without a single CVE.","tags":["supply chain","appsec"],"takeaways":[],"title":"Dependency Confusion: How Package Managers Can Become an Attack Path","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-31","url":"https://firewallsync.com/posts/dependency-confusion-how-package-managers-can-become-an-attack-path/"},{"author":"John Marshall","description":"Ivanti CVE-2026-111357 and CVE-2026-118527 are in CISA KEV and chain with CVE-2026-104935. Fixed versions, the gaps in the record, and how to triage your VPN.","faq":[{"a":"Not one we could confirm. The CVE record for CVE-2026-111357 was published by MITRE, not Ivanti, and lists only 22.8R2.4 as affected. CISA's KEV entry points to Ivanti's CVE-2026-104935 advisory and tells agencies to apply mitigations or stop using the product if mitigations are unavailable. Check Ivanti's advisory and support portal directly for the current fix status.","q":"Is there a patch for CVE-2026-111357?"},{"a":"The Ivanti-issued CVE record lists 25.1.3.0, 25.1.3.1 and 22.8R2.5 as unaffected and says 22.7R2.14 and earlier are not impacted. Rapid7's summary of the advisory also lists 25.1.1.2 as a fix for the 25.1.1.x line; confirm that against Ivanti's advisory before relying on it.","q":"Which Ivanti Connect Secure versions fix CVE-2026-104935?"},{"a":"Do not rely on it alone. Ivanti recommends running the Integrity Checker Tool regularly, but Volexity, as reported by The Hacker News, said the tool does not detect this chain, and Rapid7 quotes Ivanti saying reliable indicators of compromise are not available.","q":"Will the Ivanti Integrity Checker Tool detect this attack?"},{"a":"CVE-2026-104935 was added on 13 October with a 16 October due date. CVE-2026-111357 was added on 30 October with a 2 November due date and forensic triage set to Yes. CVE-2026-118527 was added on 30 October with a 13 November due date and forensic triage set to No.","q":"What are the KEV due dates?"}],"markdown_url":"https://firewallsync.com/posts/ivanti-connect-secure-cve-2026-111357-cve-2026-118527-kev/index.md","published":"2026-10-31","reading_minutes":7,"summary":"CISA added two more Ivanti Connect Secure flaws to KEV on 30 October: CVE-2026-111357, a privilege escalation to root, and CVE-2026-118527, a predictable-identifier bug that lets an authenticated attacker write files to a known path. Both chain with CVE-2026-104935, the authentication bypass Ivanti patched on 13 October. Here is what the public records confirm, what they do not, and what to do before the 2 November KEV deadline.","tags":["ivanti","connect secure","vpn","kev","vulnerability management","incident response"],"takeaways":["CISA added CVE-2026-111357 (privilege escalation, due 2 November) and CVE-2026-118527 (predictable identifiers enabling file writes, due 13 November) to KEV on 30 October 2026. KEV says both chain with CVE-2026-104935.","The Ivanti-issued CVE record for CVE-2026-104935 lists 25.1.3.0, 25.1.3.1, 22.8R2.5 and 22.7R2.14 as unaffected. Upgrade to one of those first: it closes the unauthenticated entry point of the chain.","We could not find a public fixed version for CVE-2026-111357 or CVE-2026-118527. The only CVE record for CVE-2026-111357 is MITRE's, and it names 22.8R2.4 as affected; CVE-2026-118527 had no published CVE record when we checked.","Volexity found the chain on a fully patched 22.8R2.4 appliance. The attacker stole LDAP credentials and used no malware, and Ivanti says reliable indicators of compromise are not available. If you ran 22.8R2.4 or earlier on the internet in October, assume compromise until you can show otherwise."],"title":"Ivanti CVE-2026-111357 and CVE-2026-118527: What to Do","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-31","url":"https://firewallsync.com/posts/ivanti-connect-secure-cve-2026-111357-cve-2026-118527-kev/"},{"author":"Craig Leslie","description":"To run user-supplied regex safely, use a linear-time engine such as RE2, cap pattern length, and isolate backtracking engines behind a timeout. Tested code.","faq":[{"a":"Regular expression denial of service. Backtracking regex engines can take exponential time when a pattern has a repeated group containing another repetition or overlapping alternatives, such as (a+)+ or (a|aa)+, and the input almost matches. OWASP's wiki lists these as 'evil' regexes and notes that if the regex itself comes from user input, 'the attacker can inject an Evil Regex'.","q":"What is ReDoS?"},{"a":"Only with an engine designed for it, or with the user's text escaped so it is matched literally. With RE2, Go's regexp or Rust's regex crate, matching time stays linear in the input. With backtracking engines such as Python's re, JavaScript's RegExp and .NET, a user-supplied pattern can hang a thread, and .NET's own docs say its engine does not protect against hostile patterns.","q":"Is it safe to build a regex from user input?"},{"a":"They limit the damage from your own patterns, but a timeout is not a complete defence against hostile ones. Ruby's Regexp.timeout and .NET's matchTimeout stop runaway backtracking. Microsoft says these features 'are not intended as a security boundary against malicious patterns', because a pattern can also consume resources at compile time or regardless of the input.","q":"Do regex timeouts prevent ReDoS?"},{"a":"Not by default. V8 has an experimental non-backtracking engine behind the --enable-experimental-regexp-engine flag, which adds a non-standard l (linear) flag, but V8 describes it as experimental and much slower than its default engine on common patterns. For untrusted patterns in Node.js, use an RE2 binding or isolate matching in a worker you can terminate.","q":"Does Node.js have a safe regex engine?"}],"markdown_url":"https://firewallsync.com/posts/user-supplied-regex-redos-safe-engines/index.md","published":"2026-10-31","reading_minutes":10,"summary":"Search filters, alert rules, routing configs and log tools all end up accepting regular expressions from users. In a backtracking engine that is a denial-of-service switch you hand to whoever can type a pattern. This guide shows which engines are built for untrusted patterns (RE2, Go, Rust), which ones say plainly that they are not (.NET), what timeouts and possessive quantifiers can and cannot do, and two tested patterns for Python and Node.js.","tags":["redos","regex","denial of service","input validation","node.js","python","appsec"],"takeaways":["A backtracking engine can take exponential time on a crafted input. In our Python 3.13 test, ^(a+)+$ took 0.009 s on 18 characters and 0.54 s on 24; each extra two characters roughly quadrupled the time.","If users supply the pattern, the attacker controls both halves of the problem. .NET's documentation says its engine 'does not offer protection against untrusted patterns' and that timeouts and NonBacktracking 'are not intended as a security boundary against malicious patterns'.","Use an engine designed for untrusted patterns: RE2 (and Go's regexp, which uses its syntax) guarantees time linear in the input; Rust's regex crate guarantees O(m * n) and limits compiled pattern size by default.","Linear-time engines drop backreferences and look-around. Reject those patterns with a clear error rather than falling back to a backtracking engine.","If you must use a backtracking engine, run the match where you can kill it: a worker thread or process with a hard deadline, a pattern length cap and a memory limit. Ruby 3.2+ and .NET have built-in timeouts; Python's re and JavaScript's RegExp do not."],"title":"User-Supplied Regex: How to Run It Without ReDoS","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-31","url":"https://firewallsync.com/posts/user-supplied-regex-redos-safe-engines/"},{"author":"FirewallSync Editorial","description":"Software supply chain attacks explained: where modern build pipelines actually break, from dependency resolution to CI runners to build-time script execution.","faq":[],"markdown_url":"https://firewallsync.com/posts/software-supply-chain-attacks-where-modern-build-pipelines-break/index.md","published":"2026-10-30","reading_minutes":4,"summary":"A build pipeline trusts a long chain of things by default: the package registry, the CI runner, every plugin, every base image. Each link is a place a supply chain attack can enter without ever touching your own code.","tags":["supply chain","appsec"],"takeaways":[],"title":"Software Supply Chain Attacks: Where Modern Build Pipelines Break","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-30","url":"https://firewallsync.com/posts/software-supply-chain-attacks-where-modern-build-pipelines-break/"},{"author":"John Marshall","description":"GitLab CVE-2026-97229 is exploited and in CISA KEV. Upgrade to 19.5.1, 19.4.3, 19.3.6 or 19.2.9 to close it and its sibling regex RCE CVEs, then hunt.","faq":[{"a":"It fixes CVE-2026-97229, but the CVE-2026-110264 record lists 19.4 before 19.4.3 as affected by a second unauthenticated regex code-execution bug. Go to 19.4.3, or 19.5.1 or later, instead of stopping at 19.4.2.","q":"Is GitLab 19.4.2 safe from CVE-2026-97229?"},{"a":"No action is needed on GitLab-managed infrastructure. GitLab's 19.4.2 release post says GitLab.com was already running the patched version and Dedicated tenants were patched before the release was published. The risk is on self-managed CE and EE instances.","q":"Is GitLab.com affected by CVE-2026-97229?"},{"a":"CISA added it to the Known Exploited Vulnerabilities catalog on 2026-10-29, which requires evidence of active exploitation, and CISA's SSVC entry in the CVE record lists Exploitation as 'active'. Neither source publishes attack details or indicators.","q":"Is CVE-2026-97229 being exploited?"},{"a":"Only if your application compiles regular expressions from strings an attacker controls, for example Regexp.new(params[:pattern]) or /#{input}/. Ruby's advisory says applications that only use regexes written by their own developers are not affected. If you do compile user patterns, upgrade to Ruby 4.0.9, 3.4.12 or 3.3.13.","q":"Do I need to update Ruby if I don't run GitLab?"}],"markdown_url":"https://firewallsync.com/posts/gitlab-cve-2026-97229-ci-regex-rce-patch-and-hunt/index.md","published":"2026-10-30","reading_minutes":8,"summary":"CISA added GitLab CVE-2026-97229, an unauthenticated remote code execution bug in how GitLab compiles regular expressions from CI/CD configuration, to KEV on October 29 with a November 1 deadline. The 19.4.2 fix is not the end of it: a sibling bug, CVE-2026-110264, needs 19.4.3 or 19.5.1. Here are the versions that close all four related CVEs, a script to check yours, GitLab's detections, and what Ruby shops outside GitLab should check.","tags":["gitlab","vulnerability management","ci/cd","ruby","incident response"],"takeaways":["CVE-2026-97229 lets an unauthenticated user run code on a self-managed GitLab server through a crafted regular expression in CI/CD configuration. GitLab rates it CVSS 3.1 10.0. CISA added it to KEV on 2026-10-29 with a due date of 2026-11-01.","GitLab fixed it in 19.4.2, 19.3.5 and 19.2.8 (October 12, 2026). Those builds are still exposed to CVE-2026-110264, another unauthenticated regex RCE fixed on October 21 in 19.5.1, 19.4.3, 19.3.6, 19.2.9, 19.1.9, 19.0.10 and 18.11.13.","The minimum safe target is the October 21 release on your branch or later. CVE-2026-110264 affects every version from 7.0, so any branch older than 18.11 has no fix and must be upgraded.","GitLab published named detections for self-managed instances, including 'Onigmo native worker crash' and 'ci regexp guard rejection'. Because KEV marks this CVE for forensic triage, check for compromise as well as patching.","The root bugs are in Ruby's regex engine, Onigmo. Ruby 4.0.9, 3.4.12 and 3.3.13 fix them. Any Ruby app that passes attacker-controlled strings to Regexp.new is in scope, and Regexp.timeout does not help."],"title":"GitLab CVE-2026-97229: Patch the Regex RCE, Then Hunt","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-30","url":"https://firewallsync.com/posts/gitlab-cve-2026-97229-ci-regex-rce-patch-and-hunt/"},{"author":"Craig Leslie","description":"Secure file upload in Node.js: Express and Multer limits, magic-byte allowlist, random names, storage outside the webroot and safe download headers, tested.","faq":[{"a":"No. OWASP's File Upload Cheat Sheet lists double extensions (.jpg.php), null bytes, case changes, alternative extensions such as .phtml, and NTFS alternate data streams as common bypasses. Ignore the user's extension entirely: detect the type from content, then store the file under a name and extension your server chooses.","q":"Is checking the file extension enough to secure an upload?"},{"a":"No. The file-type package's own readme calls detection 'a best-effort hint' that 'does not guarantee the file is actually of that type'. In our test a valid PNG with a PHP payload appended was detected as image/png. Combine detection with safe storage, safe serving and, for images, re-encoding.","q":"Is checking magic bytes enough?"},{"a":"Only if you need them, and then serve them as attachments or from a separate origin. SVG is XML text that can contain script, and magic-byte libraries such as file-type do not detect text formats. Our example allowlist rejects SVG with 415.","q":"Should I allow SVG uploads?"},{"a":"OWASP's order of preference is a different host first, then outside the webroot, and only then inside the webroot with write-only permissions. Object storage with private buckets and short-lived download URLs counts as a different host.","q":"Where should uploaded files be stored?"}],"markdown_url":"https://firewallsync.com/posts/secure-file-upload-nodejs-express-multer/index.md","published":"2026-10-30","reading_minutes":9,"summary":"Most upload bugs come from trusting something the client sent: the filename, the Content-Type, or the idea that a file sitting in a public folder won't be run. This guide builds an Express 5 and Multer 2 upload endpoint that limits size and count, checks magic bytes against an allowlist, renames and stores files outside the webroot, and serves them back safely. We ran ten attack requests against it, and one of them got through the type check, which is why the storage rules matter more.","tags":["file upload","node.js","express","appsec","owasp"],"takeaways":["Never trust the client's filename or Content-Type. Generate the stored name on the server and detect the type from the file's bytes.","Multer's limits default to Infinity for file size, file count and parts. Set fileSize, files, fields and parts on every upload route, and never mount Multer globally.","Magic-byte checks reject obvious fakes (a PHP script labelled image/jpeg got 415 in our test), but a valid PNG with PHP appended passed. Type detection is a filter, not a guarantee.","What made the polyglot harmless was storage: a random name with an extension we chose, mode 0600, in a directory no web server or interpreter serves.","Serve uploads through a handler that checks ownership and sends Content-Disposition: attachment, X-Content-Type-Options: nosniff and a sandbox CSP. Re-encoding images (we used sharp) strips appended payloads."],"title":"Secure File Uploads in Node.js: A Tested Express Setup","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-30","url":"https://firewallsync.com/posts/secure-file-upload-nodejs-express-multer/"},{"author":"John Marshall","description":"Node.js decompression bombs: maxOutputLength works for gunzipSync but streams ignore it. Tested limits for zlib, Brotli, zstd, Express and fetch().","faq":[{"a":"No. The Node.js docs say it limits output size 'when using convenience methods', meaning the one-shot functions such as zlib.gunzip() and zlib.gunzipSync(). In our test on Node.js 22.23.4, zlib.createGunzip({ maxOutputLength: 1048576 }) passed all 104,857,600 bytes through without an error.","q":"Does maxOutputLength protect zlib streams in Node.js?"},{"a":"Not with its default body parsers. In Express 5.3.0 with body-parser 2.4.0, express.json() counted the decompressed bytes against its 100 KB limit and returned 413 for 10 MB bodies sent compressed as gzip (10,232 bytes), Brotli (41 bytes) and zstd (355 bytes). Raising the limit raises your exposure, and custom middleware that pipes req through zlib needs its own cap.","q":"Is Express vulnerable to gzip bombs?"},{"a":"No. Content-Length is the compressed size. In our fetch() test the header said 101,941 bytes and the decoded body was 104,857,600 bytes. Count bytes after decompression.","q":"Is checking Content-Length enough?"},{"a":"zstd support (zstdCompress, zstdDecompress and the Zstd stream classes) was added in Node.js 22.15.0 and is marked Stability 1, Experimental, in the Node.js 22 documentation.","q":"Which Node.js versions support zstd in zlib?"}],"markdown_url":"https://firewallsync.com/posts/nodejs-decompression-bomb-zlib-limits/index.md","published":"2026-10-29","reading_minutes":7,"summary":"A 100 KB gzip body can decompress to 100 MB. Node's maxOutputLength stops that in the one-shot zlib functions, but streams silently ignore it, and fetch() decompresses responses with no cap at all. We tested where the limits hold in Node.js 22 and Express 5, with code you can copy.","tags":["node.js","denial of service","api security","appsec"],"takeaways":["A decompression bomb is small compressed input that expands into huge output. In our test, 100 MiB of zeros compressed to 101,941 bytes with gzip, 3,219 bytes with zstd and 172 bytes with Brotli.","maxOutputLength works on Node's one-shot functions (gunzipSync, brotliDecompress, zstdDecompressSync and the rest): they throw RangeError ERR_BUFFER_TOO_LARGE. Its default is buffer.kMaxLength, which is effectively no limit.","Streams ignore maxOutputLength. createGunzip({ maxOutputLength: 1 MiB }) let all 100 MiB through in our test. Put a byte-counting Transform after the decompressor instead.","Express 5's express.json() applies its 100 KB default limit after decompression, and it rejected 10 MB bombs in gzip, Brotli and zstd with 413. Keep the limit low; inflate: false rejects compressed bodies outright.","fetch() decompresses gzip responses automatically and arrayBuffer() has no size cap. Read response bodies through a capped reader whenever the URL or server isn't yours."],"title":"Node.js Decompression Bombs: zlib Limits That Work","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-29","url":"https://firewallsync.com/posts/nodejs-decompression-bomb-zlib-limits/"},{"author":"Craig Leslie","description":"Magento CVE-2026-48356 (PolyShell) is on CISA's KEV list. Install the 2026-jul builds, block pub/media/custom_options, then hunt CMS blocks and uploads.","faq":[{"a":"Yes. CISA added it to the Known Exploited Vulnerabilities catalog on October 27, 2026, and CISA's enrichment of the CVE record rates exploitation as active. Sansec has reported mass exploitation of PolyShell since March 2026. When Adobe published APSB26-73 on July 14, it said it was not aware of exploits in the wild.","q":"Is CVE-2026-48356 being exploited?"},{"a":"Do not assume so. Sansec says the fix first shipped in 2.4.9, but Adobe's CVE record lists 2.4.9 as affected and 2.4.9-2026-jul as the unaffected build. Install 2.4.9-2026-jul or later, which settles the question either way.","q":"I am on Magento 2.4.9. Am I patched?"},{"a":"No. It stops uploaded files from being served or executed through the web server, which is valuable on stores with custom configs. The upload itself still succeeds. Sansec makes the same point: only the patched release fixes the flaw.","q":"Does blocking pub/media/custom_options fix the vulnerability?"},{"a":"Per Adobe's CVE record: Adobe Commerce up to 2.4.9, 2.4.8-p5, 2.4.7-p10, 2.4.6-p15, 2.4.5-p17 and 2.4.4-p18; Magento Open Source up to 2.4.9, 2.4.8-p5, 2.4.7-p10 and 2.4.6-p15; Commerce B2B up to 1.5.3, 1.5.2-p5, 1.4.2-p10, 1.3.4-p17 and 1.3.3-p18; and Adobe Commerce Events up to 1.20.0.","q":"Which Magento versions are affected by CVE-2026-48356?"}],"markdown_url":"https://firewallsync.com/posts/magento-cve-2026-48356-polyshell-patch-and-hunt/index.md","published":"2026-10-29","reading_minutes":9,"summary":"CVE-2026-48356 is the PolyShell file-upload flaw in Adobe Commerce and Magento Open Source. CISA added it to the KEV catalog on October 27, 2026. Here are the fixed builds, why a store already on 2.4.9 may still need an update, and the checks that find an existing compromise.","tags":["vulnerability management","incident response","php","file upload","remote code execution"],"takeaways":["CVE-2026-48356 is an unrestricted file upload (CWE-434) in Adobe Commerce, Commerce B2B, Magento Open Source and Adobe Commerce Events. Adobe fixed it in bulletin APSB26-73 on July 14, 2026. CISA added it to the KEV catalog on October 27, 2026, with a federal due date of November 10.","Fixed builds are the 2026-jul releases: 2.4.9-2026-jul down to 2.4.4-2026-jul for Commerce, 2.4.6-2026-jul and later for Open Source, and Commerce Events 1.21.0. Adobe's CVE record lists plain 2.4.9 as affected, so a store on 2.4.9 should still move to 2.4.9-2026-jul.","Sansec, which named the flaw PolyShell, says it abuses file uploads through the guest-cart REST API. Files land in pub/media/custom_options/quote/. Whether that becomes code execution depends on your web server configuration.","Patching does not remove anything already planted. Check pub/media/custom_options/ for files, search cms_block and cms_page for injected scripts, and review admin accounts created since March 2026.","Adobe scores the flaw as needing user interaction (CVSS 3.1 9.3 in the CVE record). Do not let that lower its priority: it is on KEV, and Sansec reports hundreds of stores compromised through it."],"title":"Magento CVE-2026-48356 (PolyShell): Patch and Hunt","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-29","url":"https://firewallsync.com/posts/magento-cve-2026-48356-polyshell-patch-and-hunt/"},{"author":"Craig Leslie","description":"PHP session_regenerate_id done right: rotate before setting auth, keep old IDs on a timer, enforce idle and absolute timeouts. Tested code for PHP 8.3.","faq":[{"a":"PHP's manual says not to delete active session data straight away, because concurrent requests or a lost Set-Cookie response can leave a legitimate user on the old ID. Mark the old session with a timestamp and refuse it after a short grace period instead. Garbage collection removes it later.","q":"Should I call session_regenerate_id(true) to delete the old session?"},{"a":"No. Strict mode rejects IDs PHP has never issued, which stops an attacker inventing one. It does not stop an attacker who gets a real, server-issued ID by visiting your site and plants it on a victim. Regenerating the ID at login is what breaks that attack.","q":"Isn't session.use_strict_mode enough to stop session fixation?"},{"a":"Always on login and any privilege change. PHP's manual also suggests periodic rotation, giving 'every 15 minutes for security sensitive content' as an example. Periodic rotation limits how long a stolen ID stays useful, but it does not replace rotating on login.","q":"How often should PHP regenerate the session ID?"},{"a":"Not reliably. It marks data as garbage after that many seconds (1440 by default), and cleanup only runs when garbage collection happens to run. PHP's manual says developers must not rely on it for session expiry. Store your own timestamps and enforce idle and absolute timeouts in code.","q":"Does session.gc_maxlifetime log users out after 24 minutes?"}],"markdown_url":"https://firewallsync.com/posts/php-session-regenerate-id-stop-session-fixation/index.md","published":"2026-10-28","reading_minutes":9,"summary":"Calling session_regenerate_id() at login is the start of PHP session security, not the whole of it. This guide shows the fixation attack strict mode does not stop, a tested helper for timestamp-based rotation, idle and absolute timeouts, and the grace-period shortcut that quietly undoes the fix.","tags":["php","session security","authentication","appsec"],"takeaways":["session.use_strict_mode only rejects session IDs PHP never issued. An attacker who visits your site gets a real ID and can plant it on a victim. In our test the victim logged in, and the attacker's copy of that ID showed the victim's account.","Regenerate the session ID on every privilege change, before you write the authenticated user into $_SESSION. In our test, rotating at login left the attacker's ID anonymous.","Don't delete the old session immediately. PHP's manual says to timestamp it and refuse it after a short grace period, so concurrent requests and dropped cookies don't log users out.","Don't forward an old ID to the new one after login. The manual's lost-cookie workaround does that, and in our test it gave the attacker the authenticated session again. Keep forwarding for periodic rotation only.","session.gc_maxlifetime is garbage collection, not a timeout. Enforce idle and absolute timeouts yourself, with timestamps in $_SESSION."],"title":"PHP session_regenerate_id: Stop Session Fixation Right","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-28","url":"https://firewallsync.com/posts/php-session-regenerate-id-stop-session-fixation/"},{"author":"John Marshall","description":"Brotli CVE-2026-111497: fixed versions for libbrotli, Node.js, Python and Go, why version strings mislead, and a tested script that checks source for the fix.","faq":[{"a":"CISA added it to the Known Exploited Vulnerabilities catalog on October 27, 2026, and its CVE record rates exploitation as 'active'. Neither CISA nor the vendors have published details of the attacks. Wiz writes that researchers have demonstrated exploitation against unpatched Node.js.","q":"Is CVE-2026-111497 being exploited?"},{"a":"Not necessarily. Node.js applied the fix to its bundled copy without changing the version string. Node.js 22.23.4 reports Brotli 1.1.0 but contains the fix. Go by the Node.js version: 22.23.4, 24.22.1 or 26.12.1 and later are fixed.","q":"My Node.js says process.versions.brotli is 1.1.0. Am I vulnerable?"},{"a":"The flaw is in the decoder. Compressing your own responses does not decode untrusted data. Risk comes from decompressing data someone else controls: request bodies, upstream or third-party responses, or files. Patch anyway: many servers and proxies link the full library.","q":"Does this affect web servers that only compress responses?"},{"a":"Brotli to 1.2.1 and brotlicffi to 1.2.1.0. Both build the Brotli C decoder. requests, urllib3 and other HTTP clients use whichever one is installed, so updating these two packages covers those clients too.","q":"Which Python packages do I need to update?"}],"markdown_url":"https://firewallsync.com/posts/brotli-cve-2026-111497-find-and-patch-every-copy/index.md","published":"2026-10-28","reading_minutes":8,"summary":"CVE-2026-111497 is an out-of-bounds write in Google's Brotli decoder, now on CISA's KEV list. It ships inside Node.js, Python packages, Go modules, Envoy and more, and patched builds can still report the old Brotli version. Here is how to find every copy and check it for the fix.","tags":["vulnerability management","supply chain","node.js","python","remote code execution"],"takeaways":["CVE-2026-111497 is an out-of-bounds write in Google Brotli's decoder, versions 1.0.3 through 1.2.0. Fixed upstream releases are 1.0.10, 1.1.1 and 1.2.1. CISA added it to the KEV catalog on October 27, 2026, with a federal due date of November 10.","The bug is in the decoder, so the exposure is anything that decompresses Brotli data an attacker controls: request bodies, responses from URLs an attacker can influence, uploaded files.","Node.js fixed it in 22.23.4, 24.22.1 and 26.12.1, but those builds still report process.versions.brotli as 1.1.0 or 1.2.0. Check the Node.js version, not the Brotli version.","Python's Brotli (1.2.1) and brotlicffi (1.2.1.0), Go's andybalholm/brotli (v1.2.7), Envoy, Istio and Bun all shipped fixes. requests and urllib3 start advertising Brotli as soon as either Python package is installed.","For vendored copies, check the code rather than the version string. Our script below spots the fix in decode.c or decode.go and correctly classified 12 upstream, Node.js and Go sources."],"title":"Brotli CVE-2026-111497: Find and Patch Every Copy","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-28","url":"https://firewallsync.com/posts/brotli-cve-2026-111497-find-and-patch-every-copy/"},{"author":"Craig Leslie","description":"php.ini hardening: the settings PHP's production template leaves open, tested examples, where to set them so apps can't override them, and what not to rely on.","faq":[{"a":"It is a sensible base: errors are logged instead of displayed and remote includes are off. But it leaves expose_php on, strict session mode off and the session cookie's HttpOnly, Secure and SameSite flags unset. Change those for any internet-facing application.","q":"Is php.ini-production secure enough to use as-is?"},{"a":"Lax is the usual choice for a session cookie because the cookie is still sent when a user follows a link to your site from elsewhere. Strict breaks that flow: users arriving from an external link appear logged out. Neither replaces CSRF tokens for state-changing requests.","q":"Should I set session.cookie_samesite to Lax or Strict?"},{"a":"It makes the obvious commands fail. Our test with system disabled raised 'Call to undefined function system()'. But PHP's manual says the directive can be circumvented, and attackers who can run PHP code have other routes. Treat it as a speed bump and rely on OS-level isolation.","q":"Does disable_functions stop a webshell?"},{"a":"Many settings, yes, through ini_set() or .user.ini, depending on each directive's changeable mode. Settings applied with php_admin_value or php_admin_flag in a PHP-FPM pool cannot be overridden with ini_set, and INI_SYSTEM settings such as disable_functions cannot be changed at runtime at all.","q":"Can an application override settings in php.ini?"}],"markdown_url":"https://firewallsync.com/posts/php-ini-security-hardening-production-settings/index.md","published":"2026-10-27","reading_minutes":10,"summary":"PHP's own php.ini-production template still leaves several security settings at permissive defaults, including strict session mode, cookie flags and the X-Powered-By header. This guide lists the settings worth changing, shows what each one does with tested examples, explains where to set them so applications can't undo them, and is honest about the two directives that are not security boundaries.","tags":["php","hardening","session security","configuration"],"takeaways":["PHP's shipped php.ini-production is safer than no php.ini, but it is not hardened. Checked against a 13-setting baseline, the PHP 8.4.27 template still differed on 6: expose_php, allow_url_fopen, session.use_strict_mode, session.cookie_httponly, session.cookie_secure and session.cookie_samesite.","Turn on session.use_strict_mode. PHP's manual calls it 'mandatory for general session security', yet it defaults to 0. In our test, PHP accepted an attacker-chosen session ID with it off and replaced it with a fresh one with it on.","Set security settings where applications cannot undo them: php.ini or php_admin_value / php_admin_flag in the PHP-FPM pool. Values set with php_value, .user.ini or ini_set() can be changed by the application.","disable_functions and open_basedir are useful friction, not sandboxes. PHP's manual says disable_functions 'can be circumvented and should not be considered a sufficient security measure for shared hosting environments'. Isolate untrusted code with separate users, pools or containers.","Hardening does not replace patching. October's PHP2Shell flaws (CVE-2026-109618 and CVE-2026-109619) gave remote code execution on default configurations; some settings here narrow those paths, only updating closes them."],"title":"php.ini Security Hardening: Production Settings That Matter","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-27","url":"https://firewallsync.com/posts/php-ini-security-hardening-production-settings/"},{"author":"John Marshall","description":"PHP2Shell (CVE-2026-109618, CVE-2026-109619) gives unauthenticated RCE on PHP web servers. Fixed: 8.2.35, 8.3.36, 8.4.27, 8.5.12. Stopgaps and hunt steps.","faq":[{"a":"PHP 8.2.35, 8.3.36, 8.4.27 and 8.5.12, or any later release on the same branch. PHP 8.6.0 RC4 also contains the fix. PHP 8.1 and older are end of life and get no upstream release, though the PHP security team published minimal source backports for 7.4, 8.0 and 8.1.","q":"Which PHP versions fix CVE-2026-109618 and CVE-2026-109619?"},{"a":"Not necessarily. Distribution packages backport security fixes without changing the upstream version number. Ubuntu's 8.3.6-0ubuntu0.24.04.13 package, for example, carries a patch named CVE-2026-109618_109619.patch. Check your package changelog for the CVE IDs instead of comparing version strings.","q":"My server says PHP 8.3.6. Am I vulnerable?"},{"a":"PHP's advisory names the web-serving SAPIs (PHP-FPM, mod_php, php-cgi, FrankenPHP and LiteSpeed) and rates mod_php as highest risk. It does not single out the CLI. The CLI binary usually comes from the same package, so updating the package covers both.","q":"Is PHP CLI affected?"},{"a":"CISA added both CVEs to its Known Exploited Vulnerabilities catalog on October 26, 2026. Amazon reported exploitation attempts against its honeypots starting October 17, before disclosure. As of October 23, security firms running honeypots said no successful remote code execution had been confirmed in real-world attacks. Treat it as exploited and patch.","q":"Is PHP2Shell being exploited?"}],"markdown_url":"https://firewallsync.com/posts/php2shell-cve-2026-109618-cve-2026-109619-patch-and-hunt/index.md","published":"2026-10-27","reading_minutes":10,"summary":"Two PHP core flaws, CVE-2026-109618 and CVE-2026-109619, allow unauthenticated remote code execution against PHP that serves web requests. CISA added both to the KEV catalog on October 26 with a three-day deadline. Here are the fixed versions, why php -v can mislead you on distribution builds, the stopgaps PHP itself recommends, and what to hunt for.","tags":["vulnerability management","incident response","php","remote code execution"],"takeaways":["CVE-2026-109618 (RCE via out-of-memory corruption) and CVE-2026-109619 (RCE via stale global state) are flaws in PHP itself, not in a framework or CMS. Every PHP installation that serves web requests is in scope: PHP-FPM, Apache mod_php, php-cgi, FrankenPHP and LiteSpeed.","Upstream fixes are PHP 8.2.35, 8.3.36, 8.4.27 and 8.5.12, released October 21-22, 2026. PHP 8.1 and older get no upstream release; the PHP security team published source backports for 7.4, 8.0 and 8.1, and distributions ship their own patched packages.","On distribution builds, php -v is not enough. An Ubuntu 24.04 package we inspected reports PHP 8.3.6 but its changelog lists the fix for both CVEs. Check the package changelog, not the version string.","CISA added both CVEs to the KEV catalog on October 26, 2026, with a federal due date of October 29 and the forensic triage flag set. Amazon reported exploitation attempts from October 17; honeypot operators reported no confirmed successful RCE as of October 23.","If you cannot patch today, PHP's own stopgaps are memory_limit=-1 set non-overridably, a low post_max_size and session.upload_progress.enabled=0, applied before request parsing and followed by a worker restart. They reduce specific attack paths; they do not make PHP safe."],"title":"PHP2Shell CVE-2026-109618 and -109619: Patch, Then Hunt","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-27","url":"https://firewallsync.com/posts/php2shell-cve-2026-109618-cve-2026-109619-patch-and-hunt/"},{"author":"Craig Leslie","description":"Kubernetes audit log detections for exec, host-access pods, cluster-admin bindings, Secret sweeps, tokens and impersonation, tested on kube-apiserver v1.34.1.","faq":[{"a":"Start with actions that change who can do what or that reach into workloads and nodes: exec, attach and ephemeral containers; Pods requesting privileged mode, host namespaces or hostPath volumes; bindings to cluster-admin; Secret reads across all namespaces or in kube-system; service account tokens created by anything other than the kubelet; impersonation; and bursts of denied requests from one identity.","q":"What should I alert on in Kubernetes audit logs?"},{"a":"Match events whose objectRef.resource is pods and objectRef.subresource is exec, and keep only the ResponseComplete stage. In our kube-apiserver v1.34.1 test, one kubectl exec produced a get and a create event, each logged at two stages, and the command was in requestURI as command= query parameters.","q":"How do I detect kubectl exec in Kubernetes audit logs?"},{"a":"Yes. Authorization denials are logged with the authorization.k8s.io/decision annotation set to forbid and response code 403, and requests that fail validation or admission are logged with their error code. In our test a privileged Pod rejected with HTTP 422 was logged at Request level with its full spec.","q":"Does the audit log record requests the API server rejected?"},{"a":"The containers techniques line up with API actions: exec maps to Container Administration Command (T1609), new Pods to Deploy Container (T1610) and, with host access, Escape to Host (T1611), new cluster role bindings to Additional Container Cluster Roles (T1098.006), Secret reads to Container API (T1552.007), and permission or resource enumeration to Container and Resource Discovery (T1613).","q":"How do I map Kubernetes audit events to MITRE ATT\u0026CK?"}],"markdown_url":"https://firewallsync.com/posts/kubernetes-audit-log-detections-8-rules-tested/index.md","published":"2026-10-26","reading_minutes":12,"summary":"Eight detection rules for the Kubernetes audit log, written as jq filters and mapped to MITRE ATT\u0026CK: exec and debug sessions, pods asking for host access, cluster-admin bindings, Secret sweeps, minted tokens, impersonation, permission discovery and anonymous calls. We ran them against a scripted attack on kube-apiserver v1.34.1 and report what each caught, what it missed and how to tune it.","tags":["kubernetes","audit logging","detection engineering","threat hunting"],"takeaways":["Most Kubernetes attack techniques have to go through kube-apiserver, so a handful of audit log rules covers a lot: exec and ephemeral containers, pods that request host access, bindings to cluster-admin, Secret reads across namespaces or in kube-system, service account tokens minted by people, impersonation, permission discovery and anonymous requests.","Long-running requests such as exec are logged twice, at ResponseStarted and ResponseComplete, and one kubectl exec produced both a get and a create. Filter on ResponseComplete or every exec counts double.","Rejected requests are evidence too. A privileged Pod with a hostPath mount of / that the API server refused (HTTP 422) was still logged with its full spec. A detection that only looks at successful creates misses the attempt.","Several rules need request bodies: host-access pods, cluster-admin bindings and ephemeral container images are only visible at Request or RequestResponse level. Secrets and tokens should stay at Metadata, which is still enough to detect who touched them.","Exclude control-plane identities by name, not with a blanket 'system:' filter. Service accounts are system: identities too, and a stolen service account token is one of the most common ways in."],"title":"Kubernetes Audit Log Detections: 8 Rules We Tested","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-10-26","url":"https://firewallsync.com/posts/kubernetes-audit-log-detections-8-rules-tested/"},{"author":"John Marshall","description":"CVE-2026-0322 gives an authenticated attacker root on PAN-OS firewalls and Panorama via the management interface. Fixed releases, exposure checks, what to hunt.","faq":[{"a":"Not on its own. Palo Alto Networks describes it as an authenticated remote code execution flaw, and its CVSS v4.0 vectors rate the privileges required as high. The practical risk is stolen or reused administrator credentials, or an interface that other weaknesses expose. Treat it as urgent if your management interface is reachable from anywhere other than a management network.","q":"Is CVE-2026-0322 exploitable without a login?"},{"a":"The newest fixes per branch are 12.2.4, 12.1.10-h2, 11.2.16, 11.1.16-h3 and 10.2.18-h12. Older maintenance releases have their own hotfixes, such as 11.1.6-h39 or 10.2.13-h26. The full table is in the article.","q":"Which PAN-OS versions fix CVE-2026-0322?"},{"a":"No. Palo Alto Networks lists Cloud NGFW and Prisma Access as not impacted. Panorama is affected, both the virtual appliance and M-Series.","q":"Are Cloud NGFW and Prisma Access affected?"},{"a":"No. Palo Alto Networks says no known workarounds exist. Threat Prevention Threat IDs 510046 and 510047 give limited coverage, but only if management traffic is routed through a dataplane interface and decrypted so the firewall can inspect it. Restricting who can reach the management interface reduces the risk; upgrading removes it.","q":"Is there a workaround for CVE-2026-0322?"}],"markdown_url":"https://firewallsync.com/posts/pan-os-cve-2026-0322-management-interface-rce-patch-and-hunt/index.md","published":"2026-10-26","reading_minutes":11,"summary":"CVE-2026-0322 lets an authenticated attacker who can reach the PAN-OS management interface run code as root on PA-Series and VM-Series firewalls and Panorama. Palo Alto Networks has seen very limited exploitation, there is no workaround, and CISA gave federal agencies three days. Here are the fixed releases, a version checker built from the CVE record, and what to review after you patch.","tags":["vulnerability management","incident response","firewalls","code injection"],"takeaways":["CVE-2026-0322 is a code injection flaw (CWE-94) in the PAN-OS management interface. An authenticated attacker with network access to that interface can run arbitrary code as root. It affects PAN-OS on PA-Series and VM-Series firewalls and on Panorama; Cloud NGFW and Prisma Access are not affected.","Palo Alto Networks says 'very limited exploitation' has been observed. CISA added it to the KEV catalog on October 14, 2026, with a federal due date of October 17 and the forensic triage flag set.","There is no workaround. The fixes are hotfix releases on every supported branch, from 10.2 to 12.2, and the right one depends on which maintenance release you run. Check yours against the table below or the script in this article.","Authentication is not much of a barrier if the interface is reachable. Restrict management access to a dedicated management network or jump host, keep it off the internet and off any interface that carries GlobalProtect, and require MFA for administrators who can make changes.","Patch, then review: preserve logs first, then look at administrator logins, new or changed admin accounts and configuration commits since early October, starting with management access from addresses outside your management network."],"title":"PAN-OS CVE-2026-0322: Patch the Management Interface RCE","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-26","url":"https://firewallsync.com/posts/pan-os-cve-2026-0322-management-interface-rce-patch-and-hunt/"},{"author":"Craig Leslie","description":"A Kubernetes audit policy decides what kube-apiserver records. A tested six-rule policy, why rule order can leak Secrets, and the jq queries that use the log.","faq":[{"a":"Upstream kube-apiserver has none. If --audit-policy-file is not set, no audit events are logged at all. Managed services and installers ship their own policies; the Kubernetes docs point to the policy generated by the GCE configure-helper.sh script as a starting point.","q":"What is the default Kubernetes audit policy?"},{"a":"Metadata. It records who read or changed which Secret, when, from where, and whether it was allowed, without the request or response body. At Request or RequestResponse the body is logged, which for a Secret includes its base64-encoded data.","q":"Which audit level should I use for Kubernetes Secrets?"},{"a":"In order, top to bottom. The first rule whose users, groups, verbs, resources, namespaces or non-resource URLs all match the request sets its audit level, and later rules are ignored for that request. Put narrow, sensitive rules first and the broad catch-all last.","q":"How do audit policy rules get evaluated?"},{"a":"Yes, at Metadata level. In our v1.34.1 test the exec event's requestURI held each argument as a command= query parameter, for example command=cat\u0026command=%2Fetc%2Fshadow. It does not record what the command printed.","q":"Does the audit log show the command run with kubectl exec?"}],"markdown_url":"https://firewallsync.com/posts/kubernetes-audit-policy-what-to-log-and-at-what-level/index.md","published":"2026-10-25","reading_minutes":9,"summary":"A Kubernetes audit policy decides which API requests kube-apiserver records and how much of each it keeps. The first matching rule wins, so one misplaced rule can write Secret values into your logs. Here is a six-rule policy we ran against kube-apiserver v1.34.1, what each rule captured, and the jq queries that turn the log into answers.","tags":["kubernetes","audit logging","incident response","detection"],"takeaways":["kube-apiserver logs nothing unless it is started with --audit-policy-file, and a policy with zero rules is rejected. Each request is matched against the rules in order, and the first match sets its level: None, Metadata, Request or RequestResponse.","Log Secrets, ConfigMaps, TokenRequests and TokenReviews at Metadata only. In our test, putting a broad Request rule for writes above that rule wrote a new Secret's base64 value straight into the audit log.","Metadata is enough to answer most incident questions: who (user, impersonated user, source IPs), what (verb, resource, name, namespace), and whether it was allowed (response code and the authorization.k8s.io/decision annotation).","Log RBAC and admission-control changes at RequestResponse so you can see exactly which subjects and roles a binding granted. For exec, filter on the subresource, not the verb: one kubectl exec produced both get and create events in our test, and the command is in requestURI.","On managed clusters the provider controls the policy and audit export is often off by default. On EKS the audit log type is off until you enable it; on AKS, kube-audit-admin leaves out get and list events."],"title":"Kubernetes Audit Policy: What to Log, at What Level","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-10-25","url":"https://firewallsync.com/posts/kubernetes-audit-policy-what-to-log-and-at-what-level/"},{"author":"John Marshall","description":"CVE-2026-114625 lets any logged-in VeloCloud Orchestrator user act as superuser. Fixed in 5.2.3.18, 6.1.3.10 and 6.4.2.12. What to patch and what to hunt.","faq":[{"a":"Arista's CVE record lists VCO 5.2.3.18 and later in the 5.2 train, 6.1.3.10 and later in the 6.1 train, and 6.4.2.12 and later in the 6.4 train. Hosted and Dedicated VCO were patched by Arista. The record lists no 7.0 range for this CVE; 7.0 deployments should confirm their status with Arista.","q":"Which VeloCloud Orchestrator versions fix CVE-2026-114625?"},{"a":"Yes, but very little. Arista says a successful attack needs network access to the VCO web interface and a valid authenticated session on the VCO portal, and the minimum role is Enterprise Read Only. That makes every tenant account, including stale and shared ones, a potential entry point.","q":"Does an attacker need credentials to exploit CVE-2026-114625?"},{"a":"CISA added it to the Known Exploited Vulnerabilities catalog on October 23, 2026, which requires evidence of active exploitation, and CISA's SSVC entry in the CVE record marks exploitation as active. Arista's own CVE record describes the issue as externally discovered but does not detail attacks.","q":"Is CVE-2026-114625 being exploited?"},{"a":"No. The September fixes (5.2.3.16 and 6.4.2.8) are below the October fixed builds, so they are still inside the affected range for CVE-2026-114625. Only 5.2.3.18, 6.1.3.10 or 6.4.2.12 and later clear all three exploited VCO flaws in their trains, according to the CVE records.","q":"If I patched VCO for CVE-2026-93952 in September, am I safe?"}],"markdown_url":"https://firewallsync.com/posts/velocloud-orchestrator-cve-2026-114625-patch-and-hunt/index.md","published":"2026-10-25","reading_minutes":8,"summary":"CVE-2026-114625 is a SQL injection in on-premises Arista VeloCloud Orchestrator that lets any authenticated user, down to the Enterprise Read Only role, act as the built-in superuser operator. CISA added it to KEV on October 23, 2026. It is the third exploited VCO flaw since July, and builds that fixed the first two are still vulnerable.","tags":["vulnerability management","incident response","sd-wan","sql injection"],"takeaways":["CVE-2026-114625 affects on-premises VeloCloud Orchestrator (VCO) 5.2 before 5.2.3.18, 6.1 before 6.1.3.10 and 6.4 before 6.4.2.12. Arista says hosted and Dedicated VCO were affected and are already patched.","Any user with a valid VCO portal session can exploit it. The minimum role is Enterprise Read Only, the lowest built-in tenant role, and Arista says no configuration prevents the exposure.","CISA added it to the KEV catalog on October 23, 2026, with a federal due date of October 26 and the forensic triage flag set. Arista's CVE record does not itself describe the exploitation.","This is the third VCO flaw in CISA's KEV catalog since July 2026. On the 5.2, 6.1 and 6.4 trains, a VCO patched for CVE-2026-16812 in July or CVE-2026-93952 in September is still exposed unless it has reached the October builds.","Patch, then hunt: an attacker with operator rights on the orchestrator can change configuration pushed to every managed Edge, so review admin activity, outbound traffic from the VCO host and the persistence indicators Arista published in September."],"title":"VeloCloud Orchestrator CVE-2026-114625: Patch and Hunt","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-25","url":"https://firewallsync.com/posts/velocloud-orchestrator-cve-2026-114625-patch-and-hunt/"},{"author":"John Marshall","description":"Bound Kubernetes service account tokens expire and die with their Pod; Secret-based ones never expire. How to find, trace and retire legacy tokens.","faq":[{"a":"Tokens issued by the TokenRequest API do: a projected token defaults to one hour and must be at least ten minutes. Secret-based tokens of type kubernetes.io/service-account-token do not expire and stay valid until the Secret or ServiceAccount is deleted, apart from automatic invalidation of unused auto-generated ones from v1.29.","q":"Do Kubernetes service account tokens expire?"},{"a":"kube-apiserver's --service-account-extend-token-expiration flag defaults to true in the Kubernetes source. It extends the tokens the ServiceAccount admission controller injects into Pods to up to one year to avoid breaking clients that never reload the file, and adds a warnafter claim about an hour out. Using the token after warnafter is counted as stale.","q":"Why does my Pod's token show an expiry a year away?"},{"a":"List Secrets of type kubernetes.io/service-account-token with kubectl get secrets -A --field-selector type=kubernetes.io/service-account-token. Their kubernetes.io/legacy-token-last-used label shows the last day each was used, and kubernetes.io/legacy-token-invalid-since shows if the control plane has already invalidated it.","q":"How do I find legacy service account tokens in a cluster?"},{"a":"There is no central list of issued tokens. A Pod-bound token dies when its Pod is deleted. To invalidate every token for a ServiceAccount, delete and recreate the ServiceAccount, which changes its UID. Legacy Secret-based tokens are revoked by deleting the Secret.","q":"How do I revoke a bound service account token?"}],"markdown_url":"https://firewallsync.com/posts/kubernetes-service-account-tokens-bound-vs-legacy-secrets/index.md","published":"2026-10-24","reading_minutes":10,"summary":"Kubernetes Pods get short-lived, Pod-bound service account tokens from the TokenRequest API, but Secret-based tokens still never expire and stay valid until someone deletes the Secret. Here is how the two differ, why a default Pod token can stay valid for up to a year, and how to find, trace and remove the legacy tokens left in your clusters.","tags":["kubernetes","service accounts","tokens","least privilege"],"takeaways":["Since Kubernetes v1.22, Pods get service account tokens from the TokenRequest API through a projected volume. These tokens have an expiry and an audience, and stop working when their Pod or ServiceAccount is deleted.","Secret-based tokens (type kubernetes.io/service-account-token) don't expire and don't rotate. Kubernetes stopped creating them automatically in v1.24, but clusters older than that, and anyone who created one by hand, may still have them.","From v1.29, auto-generated legacy tokens unused for a year are labelled invalid, and deleted after a further year unused. Manually created token Secrets are never cleaned up for you.","By default kube-apiserver extends admission-injected Pod tokens to one year and marks them stale after about an hour. The serviceaccount_stale_tokens_total metric and the authentication.k8s.io/stale-token audit annotation tell you which clients aren't reloading.","Find legacy tokens with a field selector on the Secret type, read their legacy-token-last-used label, trace callers through audit annotations, then move each caller to kubectl create token, a projected token or workload identity before you delete the Secret."],"title":"Kubernetes Service Account Tokens: Retire Legacy Secrets","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-10-24","url":"https://firewallsync.com/posts/kubernetes-service-account-tokens-bound-vs-legacy-secrets/"},{"author":"Craig Leslie","description":"CVE-2026-110447 lets unauthenticated attackers run commands through Nextcloud's built-in Collabora CODE app before 26.4.303. Who is affected and what to hunt.","faq":[{"a":"Only if the app \"Collabora Online - Built-in CODE Server\" (app ID richdocumentscode) is installed at a version before 26.4.303. Run occ app:list and look for richdocumentscode. If you use Nextcloud Office with a separately run Collabora Online server, the CVE record says that deployment is not affected.","q":"Is my Nextcloud affected by CVE-2026-110447?"},{"a":"No. Collabora's guidance, as reported by BleepingComputer and eSecurityPlanet, is that disabling is insufficient: update to 26.4.303 or later, or remove the app. In the 26.4.302 source, proxy.php does not load Nextcloud at all, so a disabled app's file can still be requested directly if the web server serves the app directory.","q":"Is disabling the built-in CODE server app enough?"},{"a":"Unexpected PHP files in apps/richdocumentscode, reported examples being shx*.php and shrf4645.php; requests to proxy.php containing $(, %24%28 or ${IFS}; outbound calls to webhook.site; and failed sudo attempts by the web server user. Review logs back to at least September 10, 2026.","q":"What should I look for if I ran a vulnerable version?"},{"a":"CISA added CVE-2026-110447 to KEV on October 23, nine days after the CVE was published on October 14. The fix had been available since September 11 and exploitation was reported from September 10, so KEV's date says nothing about when your exposure began.","q":"Why did CISA add it to KEV so late?"}],"markdown_url":"https://firewallsync.com/posts/collabora-code-cve-2026-110447-nextcloud-patch-and-hunt/index.md","published":"2026-10-24","reading_minutes":9,"summary":"CVE-2026-110447 is an unauthenticated OS command injection in proxy.php of the Collabora Online Built-in CODE Server app for Nextcloud (richdocumentscode) before 26.4.303. It was exploited from September, a month before the CVE was published, and CISA added it to KEV on October 23. Update or remove the app, then hunt for web shells.","tags":["vulnerability management","incident response","nextcloud","command injection"],"takeaways":["Only the Nextcloud app \"Collabora Online - Built-in CODE Server\" (richdocumentscode) before 26.4.303 is affected. Standalone Collabora Online servers, including Docker and package installs used through the Nextcloud Office (richdocuments) connector, are not, per the CVE record and Collabora's advisory as reported.","The flaw is an unauthenticated OS command injection in proxy.php that fires when the bundled CODE server is not running, for example after a restart, upgrade or crash. Commands run as the web server account.","The fix (26.4.303) shipped on September 11, 2026. The CVE was published on October 14 and CISA added it to KEV on October 23 with a federal due date of October 26. Exploitation reports go back to September 10.","Update richdocumentscode to the current release or remove it. Collabora says disabling the app is not enough; the vulnerable proxy.php does not load Nextcloud before it runs.","Patching does not remove a web shell. Check apps/richdocumentscode for unexpected PHP files, search access logs for proxy.php requests carrying shell syntax, and rotate secrets the web server account could read if you find anything."],"title":"Collabora CODE CVE-2026-110447: Patch Nextcloud, Then Hunt","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-24","url":"https://firewallsync.com/posts/collabora-code-cve-2026-110447-nextcloud-patch-and-hunt/"},{"author":"Craig Leslie","description":"Kubernetes Secrets encryption at rest protects etcd data, not API readers or Pod creators. What each provider stops, how to verify it, and a config checker.","faq":[{"a":"No. The Kubernetes documentation says Secrets are stored unencrypted in etcd by default. If kube-apiserver runs without --encryption-provider-config, or the first provider for Secrets is identity, new Secrets are written in plain text. Managed services may differ, so check your provider's documentation.","q":"Are Kubernetes Secrets encrypted by default?"},{"a":"No. The API server decrypts Secrets for any authorized request, so anyone with get, list or watch on Secrets still sees the values. Least-privilege RBAC is the control for API access.","q":"Does encryption at rest stop kubectl get secret from showing values?"},{"a":"The Kubernetes documentation rates KMS v2 the strongest option and a good choice when using a third-party key management tool. It recommends against aescbc because of CBC's vulnerability to padding oracle attacks, and against aesgcm unless you automate key rotation.","q":"Which encryption provider should I use?"},{"a":"Read one Secret directly from etcd with etcdctl and check that the stored value begins with k8s:enc: followed by the provider name, for example k8s:enc:aescbc:v1:key1. Secrets written before you enabled encryption stay in plain text until rewritten.","q":"How do I know existing Secrets are encrypted?"}],"markdown_url":"https://firewallsync.com/posts/kubernetes-secrets-encryption-at-rest-what-it-protects/index.md","published":"2026-10-23","reading_minutes":9,"summary":"Kubernetes stores Secrets unencrypted in etcd unless you configure an EncryptionConfiguration. Turning it on protects etcd data, but not against anyone who can read Secrets through the API, create Pods, or, with local keys, read the control plane host. This guide maps each threat to the control that actually stops it, with a tested config checker.","tags":["kubernetes","secrets management","encryption","etcd"],"takeaways":["By default Kubernetes writes Secrets to etcd in plain text; the default identity provider gives no confidentiality. Base64 in a manifest is encoding, not encryption.","Encryption at rest is configured in an EncryptionConfiguration passed to kube-apiserver with --encryption-provider-config. The first provider listed encrypts new writes; later providers are only tried for reads.","It protects the stored data in etcd. It does nothing against a user with get, list or watch on Secrets, or a user who can create Pods in the namespace, because the API server decrypts for them.","Local-key providers (aescbc, aesgcm, secretbox) keep the key in a file on every control plane host, so a host compromise exposes it. KMS v2, stable since Kubernetes v1.29, keeps the key-encryption key outside the cluster.","Enabling encryption only affects new writes. Rewrite existing Secrets with kubectl get secrets --all-namespaces -o json | kubectl replace -f -, then confirm the k8s:enc: prefix in etcd."],"title":"Kubernetes Secrets Encryption at Rest: What It Protects","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-10-23","url":"https://firewallsync.com/posts/kubernetes-secrets-encryption-at-rest-what-it-protects/"},{"author":"John Marshall","description":"CVE-2026-21589 lets unauthenticated attackers read web-root files in Jira, Confluence, Bitbucket and other Atlassian Data Center apps. Fixes and hunting steps.","faq":[{"a":"Atlassian says affected Cloud products have been patched and Cloud customers need to take no action. The fix work is for self-managed Data Center installations and for Crucible and Fisheye.","q":"Is Atlassian Cloud affected by CVE-2026-21589?"},{"a":"No. Atlassian's CVE description says exploitation requires prior knowledge of the target file's exact name and path and does not allow enumeration. That limits blind browsing, but default file locations in a Java web application are well known.","q":"Can an attacker list directories with CVE-2026-21589?"},{"a":"No. Atlassian's Jira ticket for the flaw calls the mitigations limited and not a replacement for patching. Use them only to cover the time until you can upgrade, and restrict internet access to unpatched instances where you can.","q":"Is the WAF or RewriteValve mitigation enough?"},{"a":"Rotate anything stored in files under the application's web root that the requests could have reached. watchTowr showed that Crowd's WEB-INF/classes/crowd.properties can hold application credentials, which it used to create a user in jira-administrators. Treat database passwords, application tokens and integration secrets in configuration files the same way.","q":"What should I rotate if I find suspicious requests?"}],"markdown_url":"https://firewallsync.com/posts/atlassian-cve-2026-21589-data-center-file-read-patch-and-hunt/index.md","published":"2026-10-23","reading_minutes":7,"summary":"CVE-2026-21589 lets an unauthenticated attacker read files inside the web root of eight Atlassian Data Center products, including Jira, Confluence and Bitbucket. CISA added it to KEV on October 15. Here are the fixed versions, the interim mitigations, how to search access logs, and which secrets to rotate if a read succeeded.","tags":["vulnerability management","incident response","atlassian","path traversal"],"takeaways":["CVE-2026-21589 is an unauthenticated arbitrary file access flaw in eight self-managed Atlassian products: Bitbucket, Confluence, Jira Software, Jira Service Management, Bamboo and Crowd Data Center, plus Crucible and Fisheye. Atlassian rates it CVSS 4.0 9.3.","The attacker can read files within the web application root but must know each file's exact path; the flaw does not list directories. Atlassian Cloud is already patched.","Atlassian published the advisory on October 5, 2026. watchTowr published a root-cause analysis on October 6, exploitation attempts were reported the same day, and CISA added the CVE to KEV on October 15 with an October 18 due date for federal agencies.","Upgrade to a fixed release. Atlassian's WAF rule and Tomcat or urlrewrite mitigations are explicitly limited and not a substitute for patching.","Patching does not undo a read. Search access logs for '..' next to '/', '\\' or '::' after URL decoding, and rotate any credential held in a web-root file such as Crowd's crowd.properties if exposure is possible."],"title":"Atlassian CVE-2026-21589 File Read: Patch and Hunt","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-23","url":"https://firewallsync.com/posts/atlassian-cve-2026-21589-data-center-file-read-patch-and-hunt/"},{"author":"Craig Leslie","description":"LLM05:2025 improper output handling: why model output must be treated as untrusted input, and tested Python fixes for Markdown/HTML, SQL and shell commands.","faq":[{"a":"Prompt injection (LLM01:2025) is how an attacker steers the model. Improper output handling (LLM05:2025) is what lets that steered output cause harm when the application passes it to a browser, database, shell or another system. Fixing output handling contains the damage even when an injection succeeds.","q":"What is the difference between LLM05 and prompt injection (LLM01)?"},{"a":"If you display it as plain text, HTML-encoding it is the right control. If you render Markdown or HTML from the model, you need a sanitizer with an allowlist of tags, attributes and URL schemes, plus a decision about remote images, because an image URL can send data to another server when the browser loads it.","q":"Is escaping HTML enough when I display model output?"},{"a":"A read-only user limits writes, but a generated query can still read data the user should not see, or run expensive queries. OWASP recommends parameterized queries for any database operation involving model output. The safer design is for the model to choose from queries you wrote, filling validated parameters.","q":"Can I let a model generate SQL safely if I use a read-only database user?"}],"markdown_url":"https://firewallsync.com/posts/llm-improper-output-handling-treat-model-output-as-untrusted-input/index.md","published":"2026-10-22","reading_minutes":7,"summary":"OWASP's LLM05:2025 covers what happens when an application passes model output to a browser, database, shell or file path without validation. Because prompt injection can steer that output, it should be handled like user input. Here are the failure modes and tested Python patterns for HTML, SQL and tool calls.","tags":["llm security","owasp","xss","sql injection","python"],"takeaways":["OWASP's LLM05:2025 Improper Output Handling is about passing model output to other components without validation, sanitization or encoding. OWASP lists XSS, CSRF, SSRF, privilege escalation and remote code execution as possible results.","The reason is simple: anyone who can influence the prompt, including through a document or web page the model reads, can influence the output. OWASP's first mitigation is to treat the model like any other user.","Rendering model Markdown is a common gap. In our test, a sanitizer allowlist removed an onerror handler and a javascript: link, and an image filter dropped an image URL that could have carried data to an attacker's server.","Don't let the model write SQL or shell commands. Have it fill a typed request that you validate, then build the query or command yourself: parameterized SQL, and a fixed command per allowed action with no shell.","These controls limit damage after a prompt injection; they don't prevent the injection. Combine them with least-privilege tools, Content Security Policy and logging of model output."],"title":"LLM Output Handling: Treat Model Output as Untrusted Input","topic":{"name":"AI Security","slug":"ai-security","url":"https://firewallsync.com/categories/ai-security/"},"updated":"2026-10-22","url":"https://firewallsync.com/posts/llm-improper-output-handling-treat-model-output-as-untrusted-input/"},{"author":"John Marshall","description":"Mass assignment explained: how auto-binding lets clients set role or owner fields, OWASP API3:2023, and tested fixes in Python, Spring and Rails.","faq":[{"a":"No. IDOR (insecure direct object reference) is about which object you can access, for example someone else's account by changing an ID. Mass assignment is about which properties of an object you can write. OWASP's API Top 10 tracks them separately, as API1:2023 (object level) and API3:2023 (property level), and a single endpoint can have both.","q":"Is mass assignment the same as IDOR?"},{"a":"Either stops the attack if the model is an allowlist. Rejecting them, as Pydantic does with extra='forbid', makes probing visible in logs and catches client bugs early. Ignoring them is friendlier to older clients that send extra fields. Pick one deliberately; don't rely on a framework default you haven't checked.","q":"Should unknown fields be ignored or rejected?"},{"a":"Rails 8.0 added params.expect, which the Rails guide calls the safest and most explicit way to permit parameters. params.expect(user: [:name, :email]) replaces params.require(:user).permit(:name, :email). The guide warns that expect with an empty hash permits all current and future attributes.","q":"What changed in Rails 8 for strong parameters?"}],"markdown_url":"https://firewallsync.com/posts/mass-assignment-vulnerability-how-to-stop-clients-setting-role-admin/index.md","published":"2026-10-22","reading_minutes":7,"summary":"Mass assignment happens when a framework binds every field in a request onto an object, so a client can set fields like role, plan or owner that the UI never showed. Here is how it works, what OWASP's API Top 10 calls it now, and tested fixes in Python plus the Spring and Rails mechanisms.","tags":["api security","owasp","access control","python"],"takeaways":["Mass assignment is a framework feature turned against you: request fields are bound straight onto an object, so a client that adds role, plan or owner_id to the body can change fields it was never meant to touch.","OWASP's API Security Top 10 (2023) folds mass assignment into API3:2023 Broken Object Property Level Authorization, alongside exposing properties a user should not read.","The fix is an allowlist: a dedicated input model containing only the fields the caller may change. OWASP and Spring both warn that blocklists are fragile, because a new sensitive field is writable by default.","In our tested Flask example, the vulnerable endpoint accepted role=admin and plan=enterprise; the fixed one, using a Pydantic model with extra fields forbidden, returned HTTP 422 naming the rejected fields.","Allowlisting fields does not replace authorization. You still have to check that the caller may edit this object at all, and which fields they may edit can depend on who they are."],"title":"Mass Assignment: How to Stop Clients Setting role=admin","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-22","url":"https://firewallsync.com/posts/mass-assignment-vulnerability-how-to-stop-clients-setting-role-admin/"},{"author":"Craig Leslie","description":"Kubernetes RBAC audit: the 11 permissions the Kubernetes docs flag as privilege-escalation risks, why each is dangerous, and a tested script to find them.","faq":[{"a":"No. The Kubernetes RBAC Good Practices page says list and watch also effectively reveal Secret contents, because a list response includes the contents of every Secret returned.","q":"Is get on secrets the only dangerous Secret permission?"},{"a":"Normally RBAC stops you creating or updating a role with permissions you don't have, or binding a role whose permissions you don't have. The escalate verb on roles or clusterroles lifts the first restriction, and the bind verb on a role lifts the second. Both are documented as privilege-escalation risks.","q":"What do the escalate and bind verbs do?"},{"a":"No. Nodes and PersistentVolumes are cluster-scoped, and the RBAC documentation says permissions on cluster-scoped resources must be in a ClusterRole bound with a ClusterRoleBinding to be effective. A RoleBinding to such a ClusterRole only grants its namespaced permissions.","q":"Can a RoleBinding grant access to nodes or PersistentVolumes?"},{"a":"Use kubectl auth can-i with impersonation, for example: kubectl auth can-i list pods --namespace target --as system:serviceaccount:dev:dev-sa. The Kubernetes authorization documentation shows this pattern. It requires that you have impersonation rights.","q":"How do I check what a specific service account can do?"}],"markdown_url":"https://firewallsync.com/posts/kubernetes-rbac-audit-11-permissions-that-escalate-access/index.md","published":"2026-10-22","reading_minutes":10,"summary":"Some Kubernetes RBAC permissions look harmless but let the holder become someone more powerful: list on Secrets, create on Pods, get on nodes/proxy, and the escalate, bind and impersonate verbs among them. This guide lists the 11 the Kubernetes project flags and gives a tested script that finds them in your bindings.","tags":["kubernetes","rbac","least privilege","service accounts"],"takeaways":["The Kubernetes RBAC Good Practices page lists permissions that let a user or service account escalate privileges: reading Secrets, creating workloads, creating PersistentVolumes, nodes/proxy, the escalate, bind and impersonate verbs, CSR approval, token requests, admission webhook control and namespace modification.","Several are not what they look like. list and watch on Secrets reveal Secret contents; get on nodes/proxy reaches the kubelet API and can execute commands in pods while bypassing audit logging and admission control.","Permission to create Pods in a namespace implicitly grants access to the Secrets, ConfigMaps and volumes that can be mounted there, and to the API rights of any service account in that namespace. The documentation calls boundaries inside a namespace weak.","Wildcards grant rights to resource types that don't exist yet, and members of system:masters bypass RBAC entirely. Neither can be fixed by editing a RoleBinding.","Our script reads kubectl's JSON output and prints each subject bound to a risky permission, separating cluster-wide from namespace grants. We tested it on a sample file, not a live cluster."],"title":"Kubernetes RBAC Audit: 11 Permissions That Escalate Access","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-10-22","url":"https://firewallsync.com/posts/kubernetes-rbac-audit-11-permissions-that-escalate-access/"},{"author":"John Marshall","description":"Kubernetes NetworkPolicy default deny explained: how isolation works, why DNS breaks, the both-ends rule, selector traps and a safe rollout order.","faq":[{"a":"An empty podSelector ({}) selects every pod in the policy's namespace. That is how the default-deny examples in the Kubernetes documentation apply to the whole namespace, including pods created later.","q":"What does an empty podSelector mean in a NetworkPolicy?"},{"a":"A policy with Egress in policyTypes and no egress rules blocks all outbound connections, including DNS queries to the cluster DNS service. The Kubernetes documentation says you must add a separate policy that allows egress to your DNS service.","q":"Why did my pods lose DNS after I applied default deny?"},{"a":"Not with a dedicated field, but the control plane sets an immutable label, kubernetes.io/metadata.name, on every namespace with the namespace's name as its value. You can match that label in a namespaceSelector.","q":"Can I reference a namespace by name in a NetworkPolicy?"},{"a":"No. The Kubernetes documentation says anything TLS-related, forcing traffic through a gateway and targeting Services by name are outside the NetworkPolicy API; it suggests a service mesh, ingress controller or other tools for those.","q":"Does NetworkPolicy encrypt traffic or work at layer 7?"}],"markdown_url":"https://firewallsync.com/posts/kubernetes-networkpolicy-default-deny-safe-rollout/index.md","published":"2026-10-22","reading_minutes":8,"summary":"Kubernetes allows all pod traffic until a NetworkPolicy selects a pod. A default-deny policy flips that, but it also blocks DNS, needs allow rules on both ends of every connection and does nothing without a network plugin that enforces it. Here is how the rules combine and an order to roll it out without an outage.","tags":["kubernetes","network security","zero trust"],"takeaways":["By default every pod accepts and sends any traffic. A pod becomes isolated in a direction only when a NetworkPolicy selects it with that direction in policyTypes; after that, only traffic some policy allows gets through.","Policies only add allowances; there is no deny rule. A connection between two pods needs the source's egress and the destination's ingress to allow it, so a namespace-wide default deny means writing allow rules for both ends.","A default-deny egress policy also blocks DNS. Add an explicit allow to your cluster's DNS pods, on UDP and TCP port 53, in the same change.","NetworkPolicy has no effect unless your network plugin enforces it, and it covers TCP, UDP and SCTP only. It cannot apply a default to all namespaces, log blocked connections or block traffic from a pod's own node.","Roll out per namespace: inventory flows, apply allow rules first, then the default deny, then test, starting with egress-light namespaces. Also restrict who can relabel namespaces, because labels decide what the selectors match."],"title":"Kubernetes NetworkPolicy Default Deny: A Safe Rollout","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-10-22","url":"https://firewallsync.com/posts/kubernetes-networkpolicy-default-deny-safe-rollout/"},{"author":"Craig Leslie","description":"BeyondTrust Remote Support CVE-2026-102764 was exploited as a zero-day. Patch self-hosted appliances with BT26-05-RS, then hunt for SAWFLY and BARECLAW.","faq":[{"a":"Not to patch. BeyondTrust says SaaS instances of Remote Support and Privileged Remote Access were patched automatically; the Canadian Centre for Cyber Security and Google give October 9, 2026 as the date. Rapid7 and Google still recommend reviewing administrative changes and credentials, because one of the two known compromises was a BeyondTrust-hosted instance.","q":"Do BeyondTrust SaaS customers need to do anything?"},{"a":"Versions 24.3.1 through 26.2.3, according to the Canadian Centre for Cyber Security and Google. Appliances older than 24.3.1 must first be upgraded to a supported version and then patched. Google notes that upgrading does not remediate exposure that already happened.","q":"Which self-hosted versions can take the BT26-05-RS patch?"},{"a":"BeyondTrust says PRA shares the affected code but its testing found no way to exploit the flaw in PRA. It released a PRA patch anyway as a precaution and recommends applying it to self-hosted PRA.","q":"Is Privileged Remote Access (PRA) vulnerable?"},{"a":"Not necessarily. BeyondTrust says indicators from one victim rarely reappear at the next and that evidence kept only on the appliance may not survive or may have been manipulated. Google reports the actor deleted on-appliance logs. Use the behavioral hunts and external logs, not just file hashes.","q":"If my appliance shows none of the published indicators, is it clean?"}],"markdown_url":"https://firewallsync.com/posts/beyondtrust-remote-support-cve-2026-102764-patch-and-hunt/index.md","published":"2026-10-22","reading_minutes":7,"summary":"A BeyondTrust Remote Support authentication bypass was exploited as a zero-day from late August 2026. SaaS instances are patched; self-hosted appliances need the BT26-05-RS patch, and patching does not remove an attacker who is already in. Here is what to patch, what to hunt for and what to rotate.","tags":["vulnerability management","incident response","kev","remote access"],"takeaways":["CVE-2026-102764 lets an unauthenticated remote attacker perform file operations on a BeyondTrust Remote Support appliance. It affects versions up to and including 26.2.3 and is rated CVSS 4.0 9.3 by BeyondTrust.","It was exploited as a zero-day from August 29, 2026 against two customers, before BeyondTrust shipped a patch on October 8 and disclosed the flaw on October 15. CISA added it to KEV on October 16 with forensic triage required.","SaaS customers need no patch action. Self-hosted appliances on 24.3.1 through 26.2.3 need the BT26-05-RS patch; older appliances must be upgraded first. BeyondTrust also patched Privileged Remote Access as a precaution.","Google's threat intelligence group attributes the activity to ARMORY CASTLE, a suspected China-nexus espionage actor formerly tracked as UNC5221, which deployed a PHP web shell (SAWFLY), a tunneler (BARECLAW), a root backdoor and deleted local logs.","Patching closes the hole but does not evict an intruder. If an appliance was internet-reachable on a vulnerable version, review it for compromise, and if compromise is found, rebuild it and rotate every credential it stored or brokered."],"title":"BeyondTrust CVE-2026-102764: Patch, Then Hunt for SAWFLY","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-22","url":"https://firewallsync.com/posts/beyondtrust-remote-support-cve-2026-102764-patch-and-hunt/"},{"author":"John Marshall","description":"Plex Media Server CVE-2026-96652, -96654, -96655 and -96656 are exploited in the wild. Fixed in 1.43.3.10861. Deadlines, chain details and what to check.","faq":[{"a":"Version 1.43.3.10861. The CVE records, CISA's advisory VA-26-266-01 and Plex's October 19 forum update all give that build as the first unaffected version. Plex's original September 1 notice said 1.43.3; the later update narrowed it to 1.43.3.10861 and later.","q":"Which Plex Media Server version fixes these vulnerabilities?"},{"a":"The vulnerable code is still present until you update. Plex's update describes attacks against unpatched internet-facing servers, so a server that is not reachable from the internet is less exposed, but anyone who can reach it on your network could still try. Update regardless.","q":"Is my Plex server affected if remote access is turned off?"},{"a":"CISA set a due date of October 23, 2026 for CVE-2026-96654 and CVE-2026-96656 and November 3, 2026 for CVE-2026-96652 and CVE-2026-96655. CISA does not explain individual due dates in the catalog. The KEV entries show that only CVE-2026-96656 requires forensic triage, and CISA's enrichment rates only CVE-2026-96654 as automatable.","q":"Why does one CVE have a shorter KEV deadline than the others?"},{"a":"CVE-2026-96651, a path traversal in the same advisory that can expose the PlexOnlineToken, is not in the KEV catalog as of version 2026.10.21, and CISA's enrichment from September 23 records no known exploitation. It is fixed in the same build.","q":"Is CVE-2026-96651 also exploited?"}],"markdown_url":"https://firewallsync.com/posts/plex-media-server-cve-2026-96656-exploited-chain-patch-and-hunt/index.md","published":"2026-10-22","reading_minutes":7,"summary":"Four Plex Media Server flaws fixed in 1.43.3.10861 are now in CISA's KEV catalog, and Plex says the chain is being widely exploited after a public proof of concept. Here is what each CVE does, which deadlines apply, and what to check if your server was exposed.","tags":["vulnerability management","incident response","kev"],"takeaways":["Plex Media Server versions before 1.43.3.10861 are affected by four flaws that CISA added to its Known Exploited Vulnerabilities (KEV) catalog on October 20, 2026: two server-side request forgery (SSRF) bugs, a URL-injection bug and an arbitrary file write that can lead to code execution.","Plex's own October 19 update says the flaws are 'being actively and widely exploited', that a public proof of concept appeared on October 16, and that chained together they let an unauthenticated remote attacker run code as the Plex server process.","The fix has existed since Plex's September 1 security notice. If your server ran an older build and was reachable from the internet, Plex says to assume it may be compromised.","Updating removes the vulnerabilities, not an attacker's foothold. Plex lists unfamiliar .so or .dll files in the Codecs folder, connections to proxystream.st and high CPU while idle as signs to look for.","For federal agencies, the KEV deadlines are October 23 for CVE-2026-96654 and CVE-2026-96656 and November 3 for the two SSRF flaws. Only CVE-2026-96656 is marked for forensic triage."],"title":"Plex Media Server CVE-2026-96656 Chain: Patch and Hunt","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-22","url":"https://firewallsync.com/posts/plex-media-server-cve-2026-96656-exploited-chain-patch-and-hunt/"},{"author":"FirewallSync Editorial","description":"Webhook security: verify HMAC signatures on the raw body, reject stale or replayed deliveries, dedupe by event ID. Stripe, GitHub and Shopify compared, with tested code.","faq":[{"a":"It means the body was signed by someone holding the secret and wasn't changed. It doesn't mean the request is new: a captured request can be resent with the same valid signature. You also need a freshness check where the provider signs a timestamp, and duplicate detection by event or delivery ID.","q":"Does a valid signature mean the request is safe to process?"},{"a":"GitHub's documentation recommends using the X-GitHub-Delivery header, which is unique per event, to detect replays. It notes that a redelivery you request keeps the same X-GitHub-Delivery value, so storing processed delivery IDs also absorbs redeliveries.","q":"How does GitHub protect against webhook replay without a timestamp?"},{"a":"Stripe's libraries default to 5 minutes. Stripe warns not to set the tolerance to 0, which disables the recency check, and recommends keeping your server clock synchronized with NTP. Stripe generates a new timestamp and signature for each retry, so retries don't fail the check.","q":"What tolerance should I use for Stripe's timestamp?"},{"a":"Usually because the framework parsed the JSON before your code saw the body, so you're hashing a re-serialized copy. Stripe's documentation names whitespace changes, key reordering and encoding changes as causes. Other common causes are using the wrong secret, the wrong encoding (hex versus Base64) or, for Stripe, leaving the timestamp out of the signed string.","q":"Why does my signature check fail on legitimate requests?"}],"markdown_url":"https://firewallsync.com/posts/webhook-security-hmac-signatures-timestamps-and-replay-protection/index.md","published":"2026-10-19","reading_minutes":9,"summary":"A webhook receiver is a public URL that triggers real actions. Securing it takes three separate controls: HMAC signature verification against the raw body, a freshness check on a signed timestamp, and idempotent processing keyed on the event ID. Here is how Stripe, GitHub and Shopify differ, with tested Python verification code.","tags":["api security","appsec"],"takeaways":["A webhook endpoint accepts unsolicited POST requests from the internet and acts on them. Without verification, anyone who finds the URL can send a fake 'payment succeeded' event.","Forgery, replay and duplicate delivery are three different problems. Signatures stop forgery, a signed timestamp limits replay, and idempotent processing keyed on the event ID handles both replays inside the time window and the provider's own retries.","Verify the HMAC against the raw request bytes, before any JSON parsing, and compare with a constant-time function such as Python's hmac.compare_digest or Node's crypto.timingSafeEqual.","Providers differ: Stripe signs '\u003ctimestamp\u003e.\u003cbody\u003e' and sends t= and v1= values; GitHub sends sha256=\u003chex\u003e of the body with no timestamp; Shopify sends a Base64 digest. Copying one provider's verifier for another fails quietly.","IP allowlisting, which Stripe and GitHub both document, is a useful extra layer but not a substitute for signature checks."],"title":"Webhook Security: Signatures, Replay Protection, Idempotency","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-22","url":"https://firewallsync.com/posts/webhook-security-hmac-signatures-timestamps-and-replay-protection/"},{"author":"Craig Leslie","description":"SSVC deployer patch-priority model explained: the four decision points (Exploitation, System Exposure, Automatable, Human Impact), the Defer, Scheduled, Out-of-cycle and Immediate outcomes, the full 72-row decision table, worked examples with real CVE records, and a documentation inconsistency in Human Impact.","faq":[{"a":"SSVC stands for Stakeholder-Specific Vulnerability Categorization. It is a decision-tree method from the CERT Coordination Center (CERT/CC) that turns answers about a vulnerability and its context into a priority. Different stakeholders, such as suppliers, deployers and coordinators, get different models.","q":"What is SSVC?"},{"a":"Defer (do not act at present), Scheduled (act during regularly scheduled maintenance time), Out-of-cycle (act more quickly than usual, during the next available opportunity, working overtime if necessary) and Immediate (act immediately, focusing all resources, pausing regular operations if necessary). These are CERT/CC's descriptions.","q":"What are the deployer model's four outcomes?"},{"a":"No. It gives Exploitation, Automatable and Technical Impact for the CVEs CISA analyzes. The deployer model also needs System Exposure and Human Impact, which depend on your systems and mission, so you have to assess them yourself.","q":"Does CISA's Vulnrichment data give me the deployer priority?"},{"a":"No. CERT/CC describes Defer as not acting at present. Our reading is that you should reassess when inputs change, for example when exploitation is reported; CERT/CC's pages also note that applying a mitigation can change an input such as System Exposure.","q":"Is Defer the same as never patching?"},{"a":"CERT/CC presents it as an example deployer decision table (identified on the page as ssvc:DT_DP:1.0.0) and says organizations can customize SSVC. Treat it as a well-documented starting point and check the current version on the CERT/CC site.","q":"Is the 72-row table the official SSVC standard?"}],"markdown_url":"https://firewallsync.com/posts/ssvc-deployer-patch-priority-model-explained-exploitation-exposure-automatable-and-human-impact/index.md","published":"2026-10-18","reading_minutes":10,"summary":"The CERT/CC SSVC deployer model turns four inputs into Defer, Scheduled, Out-of-cycle or Immediate. This guide reproduces all 72 rows of its example decision table, shows which inputs CISA's public CVE data already supplies, and flags where the model's own pages disagree.","tags":["vulnerability management","process"],"takeaways":["The SSVC (Stakeholder-Specific Vulnerability Categorization) deployer model, published by CERT/CC, maps four inputs to one of four priorities: Defer, Scheduled, Out-of-cycle or Immediate. The page presents its 72-row decision table as an example model, not a mandate.","CISA's Vulnrichment data in a CVE record supplies two of the four inputs, Exploitation and Automatable. System Exposure and Human Impact depend on your environment, so you must supply them. CISA's third value, Technical Impact, is not an input to this model.","In the example table, Active exploitation never produces Defer, and Immediate occurs in only 3 of 72 rows, all with Active exploitation and Open exposure. With no exploitation evidence, the only Out-of-cycle row is Automatable Yes, Open exposure and Very High Human Impact.","Human Impact is where mistakes are likely. On the CERT/CC Human Impact page, the prose definitions and the lookup table disagree for several combinations of safety and mission impact. We use the table and show the differences.","Using real CVE records read on October 18, 2026, the same flaw lands anywhere from Scheduled to Immediate depending only on your exposure and Human Impact. A score from the vendor or CISA cannot decide that for you."],"title":"SSVC Deployer Model Explained: How Exploitation, Exposure, Automatable and Human Impact Set Patch Priority","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-18","url":"https://firewallsync.com/posts/ssvc-deployer-patch-priority-model-explained-exploitation-exposure-automatable-and-human-impact/"},{"author":"John Marshall","description":"Kubernetes MutatingAdmissionPolicy explained: stable since v1.36, policy and binding resources, ApplyConfiguration versus JSONPatch mutations, reinvocationPolicy, failurePolicy, matchConditions, exempt kinds, how it differs from mutating webhooks, and a cautious rollout, based on the v1.37 documentation.","faq":[{"a":"The Kubernetes documentation states that it is stable since v1.36 and enabled by default, and the Admission Controllers reference lists MutatingAdmissionPolicy among the plugins enabled by default in Kubernetes 1.37. The per-plugin text on that same page still mentions an older v1alpha1 prerequisite, so check your own cluster's version and the kube-apiserver settings before relying on it.","q":"Is MutatingAdmissionPolicy enabled by default?"},{"a":"The documentation offers both and does not rank them. An apply configuration describes the fields you want, as a CEL object, and is merged using server-side apply rules. A JSON Patch is a list of positional operations such as add, replace and test. Apply configuration is usually easier to read for adding fields, but it cannot modify atomic structs, maps or arrays. Choose JSON Patch when you need a precondition such as a test operation.","q":"Should I use ApplyConfiguration or JSONPatch?"},{"a":"Not through the REST API. The documentation exempts ValidatingAdmissionPolicies, ValidatingAdmissionPolicyBindings, MutatingAdmissionPolicies and MutatingAdmissionPolicyBindings from policies created that way, to prevent circular dependencies. Policies defined through manifest-based admission control can intercept them.","q":"Can a MutatingAdmissionPolicy mutate other policies?"},{"a":"Before. The Kubernetes admission-control documentation says mutating controllers run in the first phase and validating controllers in the second, and that a rejection in either phase rejects the whole request. A validating rule therefore sees the object after mutations.","q":"Does a mutating policy run before or after validating policies?"},{"a":"For mutations that can be expressed from the request object, the old object, the parameter resource and CEL, it is an alternative that removes the webhook server. It cannot make external calls, and the documentation says it is an in-process alternative without HTTP callouts. If your mutation needs outside data, keep the webhook.","q":"Does MutatingAdmissionPolicy replace mutating webhooks?"}],"markdown_url":"https://firewallsync.com/posts/kubernetes-mutatingadmissionpolicy-cel-mutations-apply-configuration-or-json-patch-and-when-a-webhook-is-still-needed/index.md","published":"2026-10-18","reading_minutes":12,"summary":"MutatingAdmissionPolicy lets the Kubernetes API server change incoming objects with CEL expressions instead of calling a webhook. This guide covers the three resources, the two patch types, reinvocation, failure behavior, what the docs do not say, and a cautious rollout.","tags":["kubernetes","process"],"takeaways":["MutatingAdmissionPolicy (MAP) is a Kubernetes API-server feature that rewrites incoming objects using expressions in the Common Expression Language (CEL), with no external webhook server. The Kubernetes documentation lists it as stable since v1.36 and enabled by default; it was first available in v1.30.","A policy does nothing until a MutatingAdmissionPolicyBinding references it. Each policy needs at least one mutation, written either as an apply configuration (merged like server-side apply) or as a JSON Patch.","reinvocationPolicy is a required field with two values, Never and IfNeeded. With IfNeeded the mutation can run more than once per request, so write it to be safe to repeat. The docs' own example guards against adding the same container twice.","Unlike ValidatingAdmissionPolicyBinding, the MAP binding has no Warn or Audit action. Its three fields are policyName, paramRef and matchResources, so a cautious rollout relies on narrow scoping and a test cluster, not a warn-only mode.","A webhook is still the tool when the decision needs data or logic that is not in the request, the parameter resource or CEL. The docs describe MAP as in-process and free of external HTTP callouts."],"title":"Kubernetes MutatingAdmissionPolicy: CEL Mutations with Apply Configuration or JSON Patch, and When a Webhook Is Still Needed","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-10-18","url":"https://firewallsync.com/posts/kubernetes-mutatingadmissionpolicy-cel-mutations-apply-configuration-or-json-patch-and-when-a-webhook-is-still-needed/"},{"author":"FirewallSync Editorial","description":"9 real JWT security mistakes — algorithm confusion, the alg:none bypass, missing revocation, and more — with the fixes that actually close each one.","faq":[],"markdown_url":"https://firewallsync.com/posts/jwt-security-9-mistakes-that-can-turn-tokens-into-an-attack-path/index.md","published":"2026-10-17","reading_minutes":5,"summary":"Most JWT vulnerabilities aren't flaws in the standard — they're implementation shortcuts that trust the token to describe its own validity. Here's what to check.","tags":["api security","authentication"],"takeaways":[],"title":"JWT Security: 9 Mistakes That Can Turn Tokens Into an Attack Path","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-17","url":"https://firewallsync.com/posts/jwt-security-9-mistakes-that-can-turn-tokens-into-an-attack-path/"},{"author":"Craig Leslie","description":"CVSS v4.0 vector string explained: Base, Threat, Environmental and Supplemental metrics, the CVSS-B, CVSS-BT, CVSS-BE and CVSS-BTE names, default values, strict ordering, and worked examples from real CVE records with scores recomputed and checked.","faq":[{"a":"It looks like CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N. The prefix names the version, and each segment is an abbreviated metric and value. FIRST's specification requires the eleven Base metrics and a fixed metric order.","q":"What does a CVSS v4.0 vector string look like?"},{"a":"CVSS-B is the Base score alone. CVSS-BTE also uses Threat metrics (such as Exploit Maturity) and Environmental metrics (such as security requirements and modified base metrics). The specification says the label should accompany any displayed numeric score so readers know which metric groups were used.","q":"What is the difference between CVSS-B and CVSS-BTE?"},{"a":"A published Base score assumes the worst case for Threat and Environmental data. The specification says unset Exploit Maturity is treated as Attacked, and unset security requirements are treated as High. If you have evidence that lowers either, a recalculated score can be lower.","q":"Why is my scanner's CVSS score higher than my own assessment?"},{"a":"Not necessarily. The versions use different metrics and different scoring methods, so the numbers are not directly comparable. Compare v4.0 with v4.0, and use the qualitative rating (Critical, High and so on) only as a rough bridge.","q":"Is a CVSS v4.0 score of 9.3 worse than a v3.1 score of 9.8?"},{"a":"No. CVSS describes severity under stated assumptions. Exploitation evidence comes from other sources, such as CISA's Known Exploited Vulnerabilities catalog or the EPSS probability model. See our comparison of CVSS, EPSS and KEV.","q":"Does a high CVSS score mean a vulnerability is being exploited?"}],"markdown_url":"https://firewallsync.com/posts/how-to-read-a-cvss-v4-0-vector-string-metric-by-metric-with-real-cve-records/index.md","published":"2026-10-17","reading_minutes":10,"summary":"A CVSS v4.0 score is only as meaningful as the vector behind it. This guide decodes every metric in a v4.0 vector string, shows how the same flaw scores 8.7 or 9.4 depending on one metric group, and explains why a published score usually assumes the worst case for threat data.","tags":["vulnerability management","tools"],"takeaways":["A CVSS v4.0 vector starts with CVSS:4.0, then lists metrics as NAME:VALUE pairs separated by slashes. The eleven Base metrics are mandatory and must appear in a fixed order. Every other metric is optional and defaults to Not Defined (X).","A score is only comparable to another score built from the same metric groups. FIRST's specification names them CVSS-B (Base), CVSS-BT (Base and Threat), CVSS-BE (Base and Environmental) and CVSS-BTE (all three).","Not Defined for Exploit Maturity (E) is scored as if the flaw were Attacked, the worst case. In our recalculation, the general vector of CVE-2026-102489 scored 8.7 with E omitted or set to A, 7.4 with E:P and 6.3 with E:U.","In a real record, CVE-2026-102489 and CVE-2026-102490 each carry two scores. CVE-2026-102489 scores 8.7 or 9.4 depending only on whether the subsequent-system impacts (SC, SI, SA) are Low or High.","A v4.0 score and a v3.1 score for the same flaw are not on one scale. CVE-2026-105207 is published as 9.3 under v4.0 and 9.8 under v3.1; both are Critical, but neither is a ranking against the other."],"title":"How to Read a CVSS v4.0 Vector String, Metric by Metric, with Real CVE Records","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-17","url":"https://firewallsync.com/posts/how-to-read-a-cvss-v4-0-vector-string-metric-by-metric-with-real-cve-records/"},{"author":"John Marshall","description":"Kubernetes ValidatingAdmissionPolicy explained: policy and binding resources, CEL expressions, Deny, Warn and Audit actions, failurePolicy, matchConditions, an image-registry allowlist and a memory-limit rule, and a warn-first rollout, based on the Kubernetes v1.37 documentation.","faq":[{"a":"No. Pod Security Admission offers three fixed levels and no custom rules. ValidatingAdmissionPolicy lets you write your own rules in CEL. They are complementary: the Kubernetes docs describe both as validating admission mechanisms, and nothing in them says to use one instead of the other.","q":"Is ValidatingAdmissionPolicy a replacement for Pod Security Admission?"},{"a":"No. The Kubernetes documentation describes validating admission policies as a declarative, in-process alternative to validating admission webhooks: the expressions run inside the API server. On v1.37 the ValidatingAdmissionPolicy admission controller appears in the documented list of plugins enabled by default. Managed Kubernetes services may differ, so check your provider.","q":"Do I need to install a webhook server to use it?"},{"a":"No. The documentation says Deny and Warn may not be used together because the combination would duplicate the failure in the API response body and in the HTTP warning headers. Deny with Audit, or Warn with Audit, are allowed combinations.","q":"Can Deny and Warn be used together in one binding?"},{"a":"Syntax errors are rejected when you create the policy. Type errors are reported in status.typeChecking but, per the docs, do not change the policy's behavior. Errors at evaluation time follow failurePolicy: Fail rejects the request, Ignore lets it through.","q":"What happens if my CEL expression has a bug?"},{"a":"No. A prefix check only tells you where the image reference points. It does not check what is in the image, whether it is vulnerable, or whether it is signed. See our guides to image scanning and to verifying signed images.","q":"Does a registry allowlist prove an image is safe?"}],"markdown_url":"https://firewallsync.com/posts/kubernetes-validatingadmissionpolicy-cel-rules-for-registry-allowlists-and-memory-limits-with-a-warn-first-rollout/index.md","published":"2026-10-17","reading_minutes":10,"summary":"ValidatingAdmissionPolicy lets the Kubernetes API server enforce custom rules written in CEL, with no webhook to run. This guide shows the policy, binding and parameter model, two worked rules that Pod Security Admission cannot express, and a rollout order that goes from Warn and Audit to Deny.","tags":["cloud security","process"],"takeaways":["ValidatingAdmissionPolicy (VAP) is a Kubernetes API-server feature that evaluates rules written in the Common Expression Language (CEL) at admission time, in-process, as an alternative to validating admission webhooks. The Kubernetes documentation lists it as stable since v1.30.","A policy has no effect on its own: you need a ValidatingAdmissionPolicy (the logic) and a ValidatingAdmissionPolicyBinding (the scope and the action). Optional parameter resources let one policy carry different values per binding.","Each binding chooses one or more actions: Deny, Warn or Audit. The docs say Deny and Warn cannot be combined, so a safe rollout is Warn plus Audit first, then Deny.","failurePolicy decides what happens when the policy itself errors. It defaults to Fail (reject the request); Ignore lets the request through. A policy that errors silently under Ignore enforces nothing.","VAP checks fields of the object in the request. Our two example rules, an image-registry prefix allowlist and required memory limits, cover gaps that Pod Security Admission does not address. Neither verifies image contents or signatures."],"title":"Kubernetes ValidatingAdmissionPolicy: CEL Rules for Registry Allowlists and Memory Limits, with a Warn-First Rollout","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-10-17","url":"https://firewallsync.com/posts/kubernetes-validatingadmissionpolicy-cel-rules-for-registry-allowlists-and-memory-limits-with-a-warn-first-rollout/"},{"author":"FirewallSync Editorial","description":"How to find and fix IDOR vulnerabilities: testing methodology, why sequential IDs make exploitation easier, and how to enforce object-level authorization correctly.","faq":[],"markdown_url":"https://firewallsync.com/posts/idor-vulnerabilities-how-to-find-and-fix-insecure-direct-object-references/index.md","published":"2026-10-16","reading_minutes":4,"summary":"IDOR is one of the simplest vulnerability classes to understand and one of the easiest to miss in testing, because it doesn't fail — it just returns the wrong person's data.","tags":["appsec","api security"],"takeaways":[],"title":"IDOR Vulnerabilities: How to Find and Fix Insecure Direct Object References","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-16","url":"https://firewallsync.com/posts/idor-vulnerabilities-how-to-find-and-fix-insecure-direct-object-references/"},{"author":"FirewallSync Editorial","description":"CVE record states explained: what PUBLISHED and REJECTED mean, what the disputed, unsupported-when-assigned and exclusively-hosted-service tags promise, real examples from CVE Services, and how to handle each in vulnerability triage.","faq":[{"a":"The CVE ID should no longer be used. The CVE Record Format's rejected record carries a required rejectedReasons field and an optional replacedBy list of CVE IDs. The CNA Operational Rules (version 4.2.0) say common reasons include determining that no vulnerability exists, using an incorrect CVE ID, a duplicate assignment, or merging several IDs. We found that the explanation text in real records varies, and replacedBy was empty in both rejected records we examined.","q":"What does it mean when a CVE is rejected?"},{"a":"Not necessarily. The CVE schema describes the disputed tag as used when one party disagrees with another party's assertion that an issue in software is a vulnerability. The record stays published, and the CNA Operational Rules require the explanation to be shown with it. Read the explanation, then check the vendor's position before deciding.","q":"Is a disputed CVE a false positive?"},{"a":"The CVE Record Format allows exactly one cna container, written by the CVE Numbering Authority that assigned the ID, and any number of adp containers, written by Authorized Data Publishers who add information. The CNA container is where the record's required content lives; an ADP container only needs provider metadata.","q":"What is the difference between a CNA and an ADP container?"},{"a":"No. The schema says 'unknown' means it is unknown or unspecified whether a version is affected, for reasons including that an investigation has not been done or that a vendor has not disclosed it. The CNA rules say CNAs should resolve unknown status within a reasonable timeframe, but a permanent unknown may be appropriate in some cases.","q":"Does a CVE with status 'unknown' for a version mean that version is safe?"}],"markdown_url":"https://firewallsync.com/posts/cve-record-states-published-rejected-and-disputed-what-each-means-for-vulnerability-triage/index.md","published":"2026-10-16","reading_minutes":10,"summary":"A CVE ID is not always a confirmed, fixable flaw. The CVE Record Format has two states, Published and Rejected, plus tags such as disputed and unsupported-when-assigned. Here is what the rules say, what real records look like, and how to handle each in a triage queue.","tags":["vulnerability management","fundamentals"],"takeaways":["The CVE Record Format defines two record states: PUBLISHED and REJECTED. A rejected record carries a required explanation (rejectedReasons) and an optional list of replacement IDs (replacedBy), so a scanner finding on a rejected CVE ID is stale data, not an open vulnerability.","'Disputed' is a tag on a published record, not a state. Under the CNA Operational Rules version 4.2.0 (effective August 25, 2026), a CVE Numbering Authority that marks a record disputed must use the tag and must include an explanation in the record. A disputed record is still published, and the tag only flags that someone disagrees.","Two other CNA tags change what you should do: unsupported-when-assigned (the affected products were already end-of-life, so a fix may never come) and exclusively-hosted-service (only a vendor-run service is affected, so there may be nothing for you to patch).","On October 16, 2026 the NVD API returned 408,789 CVE records, of which 390,296 were not rejected, which implies 18,493 (4.5%) were rejected in NVD's view. It listed 1,509 CVEs tagged disputed. Treat both as NVD counts on that day, not CVE Program statistics.","Do not treat a disputed or end-of-life tag as 'low risk'. The EPSS exploitation-probability file for October 15, 2026 gave an unsupported-when-assigned WordPress plugin flaw (CVE-2020-35235) a score of 0.18284, at the 97th percentile."],"title":"CVE Record States: Published, Rejected and Disputed, and What Each Means for Vulnerability Triage","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-16","url":"https://firewallsync.com/posts/cve-record-states-published-rejected-and-disputed-what-each-means-for-vulnerability-triage/"},{"author":"FirewallSync Editorial","description":"EPSS daily CSV columns: cve, epss (30-day exploitation probability) and percentile. What each means, model-version breaks and a tested Python scanner join.","faq":[{"a":"FIRST's Get the Data page lists a daily CSV at https://epss.empiricalsecurity.com/epss_scores-current.csv.gz (gzip-compressed, and the URL redirects, so curl needs -L), an API at https://api.first.org/data/v1/epss, and a historical archive at github.com/empiricalsec/epss_scores. No registration is required. We downloaded the CSV on October 16, 2026; it held the scores dated October 15, 2026.","q":"Where do I download EPSS scores?"},{"a":"FIRST calls the probability the model's direct output and the percentile a relative ranking of the same value. Use the probability for calculations such as expected counts, and the percentile when you want to compare a CVE with the rest of the scored population. Because scores cluster near zero, the percentile is the easier number to sanity-check a threshold against.","q":"Should I filter on the epss column or the percentile column?"},{"a":"Not necessarily. FIRST's FAQ states that 0.05 means an estimated 5% probability of observed exploitation activity in the next 30 days, that the base rate is around 2 to 3%, and that 5% ranks in roughly the top 10% of published vulnerabilities. In the October 15, 2026 file, 7.89% of CVEs scored 0.05 or higher.","q":"Does an EPSS score of 0.05 mean a CVE is unlikely to be attacked?"},{"a":"The file lists scored CVEs. In our test, two CVE IDs that were rejected in the CVE List (CVE-2024-1992 and CVE-2019-1508) were absent. A CVE that was published after the file was generated, or one the scorer does not cover, would also be missing; we did not test those cases.","q":"Why is a CVE missing from the EPSS file?"}],"markdown_url":"https://firewallsync.com/posts/epss-data-feed-columns-percentile-vs-probability-and-how-to-join-it-to-a-scanner-export/index.md","published":"2026-10-16","reading_minutes":12,"summary":"FIRST publishes EPSS scores as a free daily CSV with three columns. Here is what epss and percentile each promise, how the October 15, 2026 file is distributed, how it lines up with CISA's KEV catalog, and a tested Python script that adds both columns to a scanner export.","tags":["vulnerability management","tools"],"takeaways":["The EPSS (Exploit Prediction Scoring System) daily file has three columns: cve, epss (a probability from 0 to 1 that exploitation activity will be observed in the next 30 days) and percentile (the share of scored CVEs with the same score or lower). The file we downloaded was scored 2026-10-15T12:03:56Z by model v2026.06.15 and held 389,824 CVEs.","Scores are heavily skewed toward zero. In that file the median score was 0.00649 (0.65%) and only 4.44% of CVEs scored 0.1 or higher, so a score that looks small can still be high relative to other CVEs. FIRST says 0.10 sits around the 95th percentile; the file we downloaded puts it at about the 95.6th.","All 1,746 entries in CISA's KEV catalog (version 2026.10.15) had an EPSS score, but not all scored high: 1,470 (84.2%) were at or above the 90th percentile and 61 (3.5%) scored below 0.01. EPSS and KEV answer different questions, so neither replaces the other.","Use the daily CSV for bulk joins and the API for single-CVE lookups. FIRST says each model version changes the score scale, so a time series that crosses a version date (the latest is 2026-06-15) reflects a methodology change, not necessarily a change in the vulnerability.","Rejected CVE IDs are not in the file. A join must treat a missing score as 'no score', not as zero risk."],"title":"EPSS Data Feed Columns: epss vs percentile, Joined to Scans","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-16","url":"https://firewallsync.com/posts/epss-data-feed-columns-percentile-vs-probability-and-how-to-join-it-to-a-scanner-export/"},{"author":"FirewallSync Editorial","description":"Broken access control tops the OWASP Top 10. Real-world patterns — IDOR, missing function-level checks, privilege escalation — and how to prevent them.","faq":[],"markdown_url":"https://firewallsync.com/posts/broken-access-control-real-world-examples-and-how-to-prevent-it/index.md","published":"2026-10-15","reading_minutes":4,"summary":"Broken access control has topped the OWASP Top 10 since 2021 for a reason: it's not one vulnerability type, it's an entire category of ways authorization checks get skipped.","tags":["appsec","process"],"takeaways":[],"title":"Broken Access Control: Real-World Examples and How to Prevent It","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-15","url":"https://firewallsync.com/posts/broken-access-control-real-world-examples-and-how-to-prevent-it/"},{"author":"FirewallSync Editorial","description":"Kubernetes Pod Security Admission explained: privileged, baseline and restricted levels, enforce, audit and warn modes, namespace labels, exemptions, a dry-run rollout plan and a restricted-compliant pod example, based on the Kubernetes v1.37 documentation.","faq":[{"a":"Per the Kubernetes documentation: enforce rejects a violating pod; audit allows it but adds an annotation to the audit-log event; warn allows it but shows the user a warning. A namespace can set any or all modes, at different levels.","q":"What is the difference between enforce, audit and warn in Pod Security Admission?"},{"a":"The Kubernetes Enforcing Pod Security Standards page says the Pod Security Admission controller intends to replace the deprecated PodSecurityPolicies, and the PSA page links a migration guide for clusters moving off PodSecurityPolicy. It is not a one-to-one replacement: PSA offers three fixed levels and no custom policies.","q":"Does Pod Security Admission replace PodSecurityPolicy?"},{"a":"The documentation describes PSA as applying restrictions when pods are created, and says that labelling a namespace with a new enforce level tests existing pods and returns violations to you as warnings. We did not test this on a cluster, so confirm with a server-side dry run before you label.","q":"Will enabling enforce kill my running pods?"},{"a":"No separate install is needed on versions where it is enabled. The Kubernetes docs say it was available by default in v1.23 as a beta and has been generally available since v1.25. Managed Kubernetes services may differ in what you can configure, so check your provider's documentation.","q":"Do I need to install anything to use Pod Security Admission?"}],"markdown_url":"https://firewallsync.com/posts/kubernetes-pod-security-admission-namespace-labels-modes-and-a-safe-rollout-from-warn-to-enforce/index.md","published":"2026-10-15","reading_minutes":10,"summary":"Pod Security Admission is the Kubernetes built-in that blocks risky pod settings per namespace. This guide explains the three levels, the three modes, the labels that control them, what it does not cover, and a rollout order that does not break running workloads.","tags":["cloud security","process"],"takeaways":["Pod Security Admission (PSA) is a built-in Kubernetes admission controller that checks each new pod against one of three Pod Security Standards levels, privileged, baseline or restricted, chosen per namespace by labels. The Kubernetes documentation lists it as stable since v1.25.","Each namespace can set three independent modes: enforce rejects the pod, audit adds an audit-log annotation, and warn returns a warning to the user. Each mode takes its own level and optionally its own pinned Kubernetes version.","Enforce mode is applied to pods only, not to Deployments or Jobs. Audit and warn are applied to those workload objects too, which is why the documentation recommends turning them on first.","A namespace with no PSA labels is not secure by default: unless the cluster administrator has configured different defaults, the documented default level for all three modes is privileged, which is unrestricted.","Safe rollout order, from the Kubernetes docs: label namespaces with audit and warn first, test enforce changes with a server-side dry run, then set enforce. PSA checks pod fields only; it does not verify images, network paths or RBAC."],"title":"Kubernetes Pod Security Admission: Namespace Labels, Modes and a Safe Rollout from Warn to Enforce","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-10-15","url":"https://firewallsync.com/posts/kubernetes-pod-security-admission-namespace-labels-modes-and-a-safe-rollout-from-warn-to-enforce/"},{"author":"FirewallSync Editorial","description":"The CISA KEV JSON feed has 12 fields per entry. What dueDate, knownRansomwareCampaignUse and forensicTriage mean, plus a tested Python join to a scanner CSV.","faq":[{"a":"CISA's KEV page links the JSON feed, which we downloaded from https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json on October 15, 2026. CISA also mirrors the data, a CSV and the JSON schema in the cisagov/kev-data repository on GitHub.","q":"Where is the CISA KEV JSON file?"},{"a":"The cisagov/kev-data README says the catalog is updated whenever KEV changes, which typically happens on weekdays during US Eastern business hours when there are new or updated entries. The CISA catalog page itself does not state an update frequency.","q":"How often is the KEV JSON updated?"},{"a":"No. The schema defines Unknown as CISA lacking confirmation that the vulnerability has been used for ransomware. It can change to Known later.","q":"Does knownRansomwareCampaignUse = Unknown mean a vulnerability is not used by ransomware?"},{"a":"Not by itself. The date is the due date for the required action in the catalog; CISA's BOD 26-04 directive that the entries cite binds U.S. federal civilian agencies. Other organizations can use the date as a reference point for how urgently CISA treats the entry, then set their own policy.","q":"Is the dueDate field a deadline for my company?"}],"markdown_url":"https://firewallsync.com/posts/cisa-kev-json-feed-what-each-field-means-and-how-to-match-it-against-a-scanner-export/index.md","published":"2026-10-15","reading_minutes":9,"summary":"CISA publishes the Known Exploited Vulnerabilities catalog as a JSON file. Here is every field, what the schema does and does not promise, how due dates and the forensic triage flag behave in the data, and a tested 30-line Python script that joins the feed to a scanner CSV.","tags":["vulnerability management","tools"],"takeaways":["CISA's Known Exploited Vulnerabilities (KEV) catalog is available as JSON at a stable URL. The copy we downloaded on October 15, 2026 was catalog version 2026.10.14, released 2026-10-14T19:27:28Z, and its count field (1,745) matched the number of entries and the number of unique CVE IDs.","Each entry has 12 fields. The schema says knownRansomwareCampaignUse is \"Unknown\" when CISA lacks confirmation, so \"Unknown\" means unconfirmed, not \"not used\". In our copy 361 entries are \"Known\" and 1,384 are \"Unknown\".","dueDate is not a constant interval. Among the 128 entries added on or after June 10, 2026, the gap from dateAdded to dueDate was 3 days for 105 entries and 14 days for 23. Entries added in 2025 mostly show 21 days (226 of 245).","forensicTriage is \"Yes\" on 84 entries, all of them added between June 25 and October 14, 2026, and every one of them has a three-day due date. Three-day due dates also appear on 21 entries marked \"No\".","Matching is a join on cveID. Normalize case and whitespace, treat a CVE that is absent from KEV as \"no CISA exploitation record\" rather than \"safe\", and re-pull the file daily because CISA updates it on weekdays."],"title":"CISA KEV JSON Feed: Every Field and a Scanner Match Script","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-15","url":"https://firewallsync.com/posts/cisa-kev-json-feed-what-each-field-means-and-how-to-match-it-against-a-scanner-export/"},{"author":"FirewallSync Editorial","description":"How API rate limiting actually works: fixed window, sliding window, and token bucket algorithms compared, and where rate limiting fails in practice.","faq":[],"markdown_url":"https://firewallsync.com/posts/how-api-rate-limiting-actually-works-and-where-it-fails/index.md","published":"2026-10-14","reading_minutes":4,"summary":"Rate limiting is treated as a solved problem once it's implemented. In practice, the algorithm and the scope you choose determine whether it actually stops abuse or just adds latency to legitimate traffic.","tags":["api security","appsec"],"takeaways":[],"title":"How API Rate Limiting Actually Works—and Where It Fails","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-14","url":"https://firewallsync.com/posts/how-api-rate-limiting-actually-works-and-where-it-fails/"},{"author":"FirewallSync Editorial","description":"How to enforce cosign signatures in Kubernetes with Sigstore policy-controller: namespace opt-in, a keyless ClusterImagePolicy pinned to a GitHub Actions workflow, warn-mode rollout, no-match-policy and webhook failure-policy risks.","faq":[{"a":"No. Its documentation says the controller validates only namespaces that opt in with the label policy.sigstore.dev/include=true. It also describes switching to opt-out by editing the webhook configurations' matchExpressions, and warns that this may cause problems in system namespaces.","q":"Does policy-controller check every namespace by default?"},{"a":"The documentation says the behavior is set by the no-match-policy key in the config-policy-controller ConfigMap, with values warn, allow or deny, and that an image matching no policy is rejected if the key is not configured.","q":"What happens to an image that no ClusterImagePolicy matches?"},{"a":"Yes. The documentation says spec.mode set to warn admits a non-compliant image and returns a warning to the caller instead of blocking. The project also ships a policy-tester tool that evaluates a ClusterImagePolicy against an image without a cluster, per its README.","q":"Can I try a policy without blocking deployments?"},{"a":"Admission webhooks are called when requests such as creates and updates reach the Kubernetes API server, according to Kubernetes' documentation on admission webhooks. We found nothing in policy-controller's documentation describing re-verification of running pods, so treat existing workloads as unchecked until they are redeployed.","q":"Does admission-time verification protect pods that are already running?"}],"markdown_url":"https://firewallsync.com/posts/verify-signed-container-images-in-kubernetes-sigstore-policy-controller-keyless-policy-and-rollout/index.md","published":"2026-10-14","reading_minutes":9,"summary":"Signing images only helps if something refuses the unsigned ones. This guide covers Sigstore's policy-controller admission webhook: opting namespaces in, writing a ClusterImagePolicy that pins a GitHub Actions workflow identity, rolling out in warn mode, and the failure-policy trade-off the Helm chart defaults create.","tags":["supply chain","cloud security"],"takeaways":["Sigstore's policy-controller is a Kubernetes admission webhook that validates cosign signatures and attestations on container images. Per its documentation it only checks namespaces that carry the label policy.sigstore.dev/include: \"true\".","A ClusterImagePolicy lists image globs and authorities. Every matching policy must pass, and within one policy any single passing authority is enough, so one permissive authority can undo a strict one.","For keyless signatures the policy must pin an issuer and a subject. For GitHub Actions the issuer is https://token.actions.githubusercontent.com and, per Sigstore's CI quickstart, the signer identity looks like https://github.com/ORG/REPO/.github/workflows/FILE@refs/heads/BRANCH.","Policy-controller's documentation says that an image matching no policy is rejected unless the no-match-policy setting says otherwise. Set that deliberately before enabling a namespace.","The Helm chart's default values on its main branch set webhook.failurePolicy to Fail and replicaCount to 1. Under Kubernetes' Fail semantics, an unreachable webhook rejects matching requests, so plan replicas and test the failure case before enforcing."],"title":"Verify Signed Container Images in Kubernetes with Sigstore policy-controller: Keyless Policy and Safe Rollout","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-10-14","url":"https://firewallsync.com/posts/verify-signed-container-images-in-kubernetes-sigstore-policy-controller-keyless-policy-and-rollout/"},{"author":"FirewallSync Editorial","description":"CISA Vulnrichment adds Exploitation, Automatable and Technical Impact to CVE records. Pull them with curl and jq, then map them to Track, Track*, Attend or Act.","faq":[{"a":"In the containers.adp array of the CVE record (CVE Record Format 5.x), in the entry whose providerMetadata.shortName is CISA-ADP. The SSVC result is in that entry's metrics array, in an object with other.type equal to ssvc. We confirmed this on live records from the CVE Services API on October 14, 2026.","q":"Where is the CISA SSVC data in a CVE record?"},{"a":"Per the SSVC project, Automatable answers whether an attacker can reliably automate creating exploitation events for the vulnerability. Yes means steps 1 to 4 of the kill chain (reconnaissance, weaponization, delivery, exploitation) can be reliably automated. If nothing is known, the SSVC documentation says to assume Yes.","q":"What does Automatable mean in CISA's SSVC data?"},{"a":"No. Per CISA's README, Vulnrichment covers new and recent CVEs that CISA assesses; absence of the container means only that no CISA SSVC data is attached to that record. Fall back to the other signals, such as KEV listing, EPSS and the vendor advisory.","q":"Does a missing CISA-ADP entry mean a CVE is low risk?"},{"a":"No. CVSS scores the intrinsic characteristics of a flaw. SSVC is a decision tree whose outputs are actions (Track, Track*, Attend, Act). CISA's Vulnrichment adds both: SSVC decision points to every record it analyzes and, for some, a CVSS base score where the CNA (the CVE Numbering Authority that assigned the CVE) did not supply one.","q":"Is SSVC the same as CVSS?"}],"markdown_url":"https://firewallsync.com/posts/cisa-vulnrichment-how-to-read-ssvc-values-in-a-cve-record-and-turn-them-into-a-patch-decision/index.md","published":"2026-10-14","reading_minutes":10,"summary":"CISA adds three SSVC fields (Exploitation, Automatable, Technical Impact) to many CVE records. Here is where they sit in the JSON, how to pull them with one jq command, what each value means, and how CISA's published decision table turns them into Track, Track*, Attend or Act.","tags":["vulnerability management","process"],"takeaways":["CISA's Vulnrichment project writes its analysis into an Authorized Data Publisher (ADP) container inside the public CVE record. Every CVE CISA analyzes carries three SSVC decision points: Exploitation, Automatable and Technical Impact.","Those three values are a snapshot. Each carries its own timestamp, and Exploitation describes the state at analysis time, so check the timestamp before relying on it.","The three values alone do not produce a decision. CISA's decision table (Coordinator tree v2.0.3) also needs a fourth input, Mission and Well-being (low, medium or high), which depends on your organization and the system affected.","In that table the same vulnerability can land on Track, Track*, Attend or Act depending only on that fourth input. For example, an actively exploited, non-automatable, total-impact flaw is Track at low, Attend at medium and Act at high.","Vulnrichment is CISA's assessment, not a patch deadline. Binding Operational Directive deadlines apply to federal agencies for KEV entries; everyone else decides their own timelines."],"title":"CISA Vulnrichment SSVC: Read CVE Record Values with jq","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-14","url":"https://firewallsync.com/posts/cisa-vulnrichment-how-to-read-ssvc-values-in-a-cve-record-and-turn-them-into-a-patch-decision/"},{"author":"FirewallSync Editorial","description":"The most common API security misconfigurations — verbose errors, permissive CORS, exposed debug endpoints — and the fixes that actually close them.","faq":[],"markdown_url":"https://firewallsync.com/posts/the-most-common-api-security-misconfigurations-and-how-to-fix-them/index.md","published":"2026-10-13","reading_minutes":4,"summary":"API misconfigurations don't usually come from a missing feature. They come from a default that was never revisited once the API left the prototype stage.","tags":["api security","appsec"],"takeaways":[],"title":"The Most Common API Security Misconfigurations and How to Fix Them","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-13","url":"https://firewallsync.com/posts/the-most-common-api-security-misconfigurations-and-how-to-fix-them/"},{"author":"FirewallSync Editorial","description":"How to sign container images with cosign keyless signing in GitHub Actions and verify them with --certificate-identity and --certificate-oidc-issuer, plus what the August 2026 cosign verify-blob advisory (GHSA-fx35-mq7g-6g98) means for you.","faq":[{"a":"Not with keyless signing. Cosign's documentation describes keyless signing as associating identities, rather than keys, with a signature: it uses an ephemeral key, the private key is destroyed shortly after signing, and Fulcio issues a short-lived certificate binding that key to an OIDC identity. Cosign also supports signing with your own key pair via the --key flag.","q":"Do I need to manage a private key to sign container images with cosign?"},{"a":"Cosign's signing documentation says signatures use the OCI 1.1 referrer specification, and you can inspect what is attached to an image with cosign tree. The page we read did not describe the older tag-based storage, so check the documentation for the version you run.","q":"Where does cosign store the signature for a container image?"},{"a":"Cosign's command reference says that for keyless flows either --certificate-identity or --certificate-identity-regexp must be set, and either --certificate-oidc-issuer or --certificate-oidc-issuer-regexp must be set. For tokens issued to GitHub Actions jobs, the issuer is https://token.actions.githubusercontent.com.","q":"What must I pass to cosign verify for a keyless signature?"},{"a":"Only if you used cosign verify-blob or verify-blob-attestation with a legacy JSON --bundle on cosign 2.6.4 or earlier, or 3.1.2 or earlier. The advisory states that the protobuf bundle format and OCI image verification are not affected. The fixed versions are 2.6.5 and 3.1.3.","q":"Is my image-signing setup affected by the August 2026 cosign advisory?"}],"markdown_url":"https://firewallsync.com/posts/cosign-container-image-signing-keyless-signing-in-ci-and-why-verification-must-pin-an-identity/index.md","published":"2026-10-13","reading_minutes":9,"summary":"Signing a container image with Sigstore's cosign takes one command; the security comes from what the verifier insists on. Here is how keyless signing works in a GitHub Actions job, what to pin when verifying, and what an August 2026 cosign advisory shows about trusting the verifier.","tags":["supply chain","appsec"],"takeaways":["Keyless signing with cosign uses a short-lived certificate from Sigstore's Fulcio certificate authority tied to an OpenID Connect (OIDC) identity, so a CI job can sign without a long-lived private key.","In GitHub Actions the signing job needs the permission id-token: write; the official installer example signs image references of the form tag@digest, not bare tags.","A signature proves little unless the verifier pins who must have signed. For keyless flows cosign requires you to set --certificate-identity (or its regexp variant) and --certificate-oidc-issuer (or its regexp variant).","Advisory GHSA-fx35-mq7g-6g98 (published August 6, 2026, High, CVSS 7.4) affected cosign verify-blob and verify-blob-attestation with legacy JSON bundles, in cosign 2.6.4 and earlier and 3.1.2 and earlier. The advisory says OCI image verification is not affected.","Signing and verifying prove who published an image, not that it is free of vulnerabilities. Pair them with scanning, SBOMs and provenance."],"title":"Cosign Container Image Signing: Keyless Signing in CI and Why Verification Must Pin an Identity","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-13","url":"https://firewallsync.com/posts/cosign-container-image-signing-keyless-signing-in-ci-and-why-verification-must-pin-an-identity/"},{"author":"FirewallSync Editorial","description":"VEX (Vulnerability Exploitability eXchange) explained: CISA's four statuses and five not-affected justifications, how OpenVEX and CSAF encode them, and how to decide whether to trust a VEX statement when triaging scanner findings.","faq":[{"a":"Vulnerability Exploitability eXchange. CISA's June 2022 status-justification guidance describes it as a form of security advisory that lets a supplier or other party assert the status of specific vulnerabilities in a particular product.","q":"What does VEX stand for?"},{"a":"No. An SBOM is an inventory of components. A VEX document is a set of statements about the exploitability status of specific vulnerabilities in a specific product. They are often used together: the SBOM tells a scanner what to look for, and VEX tells a reader which findings the supplier says do not apply.","q":"Is VEX the same as an SBOM?"},{"a":"CISA's guidance says VEX statements are assertions the consumer may accept or reject. A reasonable approach, which is our suggestion rather than a CISA requirement, is to accept statements from suppliers you trust for the exact product version and component you run, to prefer machine-readable justifications over free text, and to keep the original finding recorded alongside the reason it was suppressed.","q":"Should I suppress a scanner finding because a vendor says 'not affected'?"},{"a":"This article covers two: OpenVEX, a lightweight JSON format whose specification says it aims to meet CISA's minimum requirements, and CSAF 2.0, an OASIS Standard that includes a VEX profile (Profile 5). CycloneDX can also carry VEX data; we did not review its specification for this article.","q":"Which VEX formats exist?"}],"markdown_url":"https://firewallsync.com/posts/vex-explained-what-a-not-affected-statement-means-and-how-to-read-one/index.md","published":"2026-10-13","reading_minutes":9,"summary":"A VEX document is a supplier's signed-off claim about whether a specific product is affected by a specific vulnerability. Here is what the four statuses and five justifications mean in CISA's guidance, how OpenVEX and CSAF express them, and why a VEX claim is something to evaluate, not obey.","tags":["supply chain","vulnerability management"],"takeaways":["VEX (Vulnerability Exploitability eXchange) lets a supplier or other party state whether a particular product is affected by a particular vulnerability. CISA's guidance lists four statuses: NOT AFFECTED, AFFECTED, FIXED and UNDER INVESTIGATION.","A NOT AFFECTED claim may carry one of five CISA justification codes, such as component_not_present or inline_mitigations_already_exist. Under both OpenVEX and CSAF, a not-affected statement must come with a machine-readable justification or a free-text impact statement.","CISA's guidance says VEX fields are assertions made by the document's author, and the consumer may accept them or not. It describes statuses as a way to inform decisions, not a discussion-ending declaration.","An SBOM says what is in a product; VEX says whether a listed vulnerability matters in that product. They are complementary documents, not substitutes.","Of the five justifications, component_not_present is the easiest for a consumer to check; inline_mitigations_already_exist is the hardest to prove, and the OpenVEX specification says so."],"title":"VEX Explained: What a \"Not Affected\" Statement Means and How to Read One","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-13","url":"https://firewallsync.com/posts/vex-explained-what-a-not-affected-statement-means-and-how-to-read-one/"},{"author":"FirewallSync Editorial","description":"A production checklist for securing a public REST API: authentication, rate limiting, input validation, and the access control gaps that get exploited most.","faq":[],"markdown_url":"https://firewallsync.com/posts/how-to-secure-a-public-rest-api-a-production-checklist/index.md","published":"2026-10-12","reading_minutes":4,"summary":"Most public API breaches don't come from a novel attack technique. They come from a well-known control that was never actually implemented before launch.","tags":["api security","appsec"],"takeaways":[],"title":"How to Secure a Public REST API: A Production Checklist","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-12","url":"https://firewallsync.com/posts/how-to-secure-a-public-rest-api-a-production-checklist/"},{"author":"FirewallSync Editorial","description":"CISA BOD 26-04 explained: who it applies to, how the four risk inputs set remediation timelines, the forensic triage requirement, the phase dates, and which practices non-federal teams can borrow.","faq":[{"a":"No. Per CISA's directive page it applies to Federal Civilian Executive Branch agency assets. Contractors are generally excluded unless a governing procurement contract directs otherwise. Private organizations can voluntarily borrow its approach.","q":"Does BOD 26-04 apply to private companies?"},{"a":"CISA's page states that BOD 26-04 revokes BOD 22-01 and BOD 19-02 and sets remediation timelines from a table keyed to KEV status, public exposure, automatability and technical impact. The table is published as an image in the directive and as Appendix A; we could not read its individual rows as text, so this article does not list every row.","q":"What replaced the old KEV deadlines?"},{"a":"A rapid assessment of whether a system has already been compromised, carried out alongside the three-day fix. CISA's implementation guidance gives baseline steps, from scoping in the first two hours through an escalation decision at 48 to 72 hours, and says the directive requires adequate triage but not those specific timings.","q":"What is forensic triage under BOD 26-04?"},{"a":"It can. KEV is based on evidence of exploitation, not on age. Five CVEs published between 2015 and 2023 were added to KEV on October 8, 2026.","q":"Does an old CVE still matter if it was disclosed years ago?"}],"markdown_url":"https://firewallsync.com/posts/cisa-bod-26-04-explained-kev-deadlines-forensic-triage-and-what-private-teams-can-borrow/index.md","published":"2026-10-12","reading_minutes":9,"summary":"CISA's Binding Operational Directive 26-04, issued June 10, 2026, replaces fixed patch deadlines for U.S. federal agencies with risk-based timelines and a three-day forensic triage requirement. A batch of five years-old flaws added to the KEV catalog on October 8, 2026 shows it in action.","tags":["vulnerability management","process"],"takeaways":["BOD 26-04, \"Prioritizing Security Updates Based on Risk,\" is a Binding Operational Directive for U.S. Federal Civilian Executive Branch agencies. CISA's page is dated June 10, 2026, and states that it revokes BOD 22-01 and BOD 19-02. It does not legally bind private companies.","Remediation time is set by four inputs: whether the flaw is in the Known Exploited Vulnerabilities (KEV) catalog, whether the asset is publicly exposed, whether exploitation is automatable, and whether the technical impact is total or partial control.","Where the timeline table requires forensic triage, agencies must remediate or mitigate within three days and also assess whether the system is already compromised.","On October 8, 2026, CISA added five older vulnerabilities (CVE IDs from 2015, 2016, 2021 and 2023) to KEV with a due date of October 11, 2026, three days later.","Age and CVSS score did not decide those additions: one entry has a National Vulnerability Database base score of 4.9."],"title":"CISA BOD 26-04 Explained: KEV Deadlines, Forensic Triage and What Private Teams Can Borrow","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-12","url":"https://firewallsync.com/posts/cisa-bod-26-04-explained-kev-deadlines-forensic-triage-and-what-private-teams-can-borrow/"},{"author":"FirewallSync Editorial","description":"SLSA Build L1, L2 and L3 explained: what provenance attestations prove, what they cannot, how Sigstore signing works, and how to verify GitHub artifact attestations with gh attestation verify.","faq":[{"a":"An SBOM lists what is inside a piece of software. SLSA provenance records how and where a specific artifact was built. They answer different questions and are often published together. Our article on the 2026 SBOM minimum elements covers the inventory side.","q":"What is the difference between SLSA and an SBOM?"},{"a":"No. SLSA says Build L3 focuses on tampering during the build and requires a hardened build platform. It does not assess the source code, its dependencies' vulnerabilities, or whether the people who wrote it are trustworthy.","q":"Does SLSA Build Level 3 mean software is secure?"},{"a":"In Sigstore's keyless flow, the signing client generates a new key pair for each signing, a certificate authority called Fulcio issues a short-lived certificate tied to the signer's OpenID Connect identity, and the private key is discarded after signing. The signature and certificate are recorded in the Rekor transparency log.","q":"What is keyless signing?"},{"a":"GitHub's documentation says attestations for private repositories use GitHub's own Sigstore instance, which has no public transparency log, while public repositories use the Sigstore Public Good Instance. The page we read does not state plan requirements, so check GitHub's current documentation for your plan.","q":"Are GitHub artifact attestations available for private repositories?"}],"markdown_url":"https://firewallsync.com/posts/slsa-build-levels-and-provenance-attestations-what-they-prove-and-what-they-dont/index.md","published":"2026-10-12","reading_minutes":8,"summary":"SLSA Build Level 1, 2 and 3 describe how trustworthy a build's provenance is, not whether the code is safe. Here is what each level promises, how Sigstore and GitHub artifact attestations fit in, and how to verify an attestation with the GitHub CLI.","tags":["supply chain","appsec"],"takeaways":["SLSA (Supply-chain Levels for Software Artifacts) Build levels L1 to L3 describe how much you can trust a build's provenance record: L1 provenance exists, L2 a hosted platform signs it, L3 the platform is hardened against tampering during the build.","Provenance says where and how an artifact was built. It does not say the source code is free of vulnerabilities or malicious changes.","SLSA v1.2 is the current version and adds a Source track; the Build track levels described here are unchanged in outline from v1.1.","GitHub states that artifact attestations alone provide SLSA v1.0 Build Level 2, and that Build Level 3 requires building in a reusable workflow.","An attestation only protects you if the consumer verifies it against an expected identity, for example with gh attestation verify and a pinned repository or signer workflow."],"title":"SLSA Build Levels and Provenance Attestations: What They Prove and What They Don't","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-12","url":"https://firewallsync.com/posts/slsa-build-levels-and-provenance-attestations-what-they-prove-and-what-they-dont/"},{"author":"FirewallSync Editorial","description":"API keys, OAuth, and JWTs solve different problems. A practical comparison of what each is actually for, and where teams commonly pick the wrong one.","faq":[],"markdown_url":"https://firewallsync.com/posts/api-authentication-api-keys-vs-oauth-vs-jwt/index.md","published":"2026-10-11","reading_minutes":4,"summary":"These three get compared as competing options, but they solve different problems. Picking the wrong one for your actual use case is where most API auth trouble starts.","tags":["api security","authentication"],"takeaways":[],"title":"API Authentication: API Keys vs OAuth vs JWT","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-11","url":"https://firewallsync.com/posts/api-authentication-api-keys-vs-oauth-vs-jwt/"},{"author":"FirewallSync Editorial","description":"CISA's 2026 SBOM Minimum Elements list 17 data fields, ten new, including component hash and license. What changed from NTIA 2021, Coverage, SPDX and CycloneDX.","faq":[{"a":"The document says the minimum elements do not create new requirements and that it is not advice for compliance, regulatory or legal purposes. It is guidance on what an SBOM should contain and how organizations should request and generate one. Whether a specific contract, procurement rule or regulation requires it is a separate question to check for your situation.","q":"Is the 2026 SBOM minimum elements guidance a legal requirement?"},{"a":"Per the 2026 document, the SBOM Author is the entity that creates the SBOM data (the entity operating the tool, not the tool itself). The Component Producer is the entity that created or developed the component. They can be the same organization or different ones, for example when a customer generates an SBOM for software someone else made.","q":"What is the difference between the SBOM Author and the Component Producer?"},{"a":"The document names both as the two data formats currently widely used and says producers may choose based on organizational or sector factors. It says organizations should accept any widely used, interoperable, machine-processable format, and should avoid accepting SBOMs for new software generated in deprecated versions of any format.","q":"Should an SBOM be SPDX or CycloneDX?"},{"a":"No. The guidance describes an SBOM as an inventory of components. Vulnerability status is communicated through separate security advisories such as VEX (Vulnerability Exploitability eXchange) and CSAF (Common Security Advisory Framework) documents, which can be matched against the components an SBOM lists.","q":"Does an SBOM list vulnerabilities?"},{"a":"The 2021 Depth element was replaced by Coverage. The 2026 document says an SBOM should include all components that make up the target software, including transitive dependencies, and that there is no minimum depth.","q":"What happened to the Depth element?"}],"markdown_url":"https://firewallsync.com/posts/sbom-minimum-elements-2026-what-cisa-and-partners-changed-from-the-2021-ntia-baseline/index.md","published":"2026-10-11","reading_minutes":9,"summary":"On July 29, 2026, CISA and international partners published updated minimum elements for a software bill of materials, replacing the 2021 NTIA document. Here is what changed in the data fields and practices, what did not, and what to ask your suppliers and your tooling for.","tags":["supply chain","appsec"],"takeaways":["The 2026 Minimum Elements for a Software Bill of Materials (SBOM) was published on July 29, 2026 by CISA with the NSA, the FBI and other national agencies. It updates and replaces NTIA's 2021 document (version 2.1; the 2025 version was a public-comment draft).","The 2021 document listed seven baseline data fields. The 2026 document's data-field table lists 17, and its own Notable Updates section names ten as new, including component hash, component license, SBOM tool name and SBOM generation context.","The guidance says it does not create new requirements; it refines how organizations should generate and request SBOMs. It is not legal or compliance advice.","Depth is replaced by Coverage: an SBOM should include all components, including transitive dependencies, with no minimum depth. Unknown data must be marked as unknown or withheld, not left blank.","SPDX and CycloneDX are named as the two widely used machine-processable formats. SWID tags, listed in 2021, were removed from the list."],"title":"SBOM Minimum Elements 2026: 17 Fields and What Changed","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-11","url":"https://firewallsync.com/posts/sbom-minimum-elements-2026-what-cisa-and-partners-changed-from-the-2021-ntia-baseline/"},{"author":"FirewallSync Editorial","description":"What HSTS, X-Content-Type-Options, Referrer-Policy and Permissions-Policy do, the values OWASP recommends, the HSTS preload requirements, and which old headers to stop sending.","faq":[{"a":"OWASP's HTTP Headers Cheat Sheet recommends, among others, Strict-Transport-Security, X-Content-Type-Options: nosniff, Referrer-Policy, a Content-Security-Policy and a restrictive Permissions-Policy. Which ones matter most depends on the site. Pages that serve HTML benefit most; a header such as X-Frame-Options does nothing for redirects or JSON API responses, according to OWASP.","q":"Which security headers should every website send?"},{"a":"No. MDN says the host must send Strict-Transport-Security over HTTPS only and that browsers ignore the header if it arrives over HTTP, so a man-in-the-middle cannot alter it. Your HTTP-to-HTTPS redirect must not include the header.","q":"Can I set HSTS over plain HTTP?"},{"a":"hstspreload.org says inclusion cannot easily be undone. Removal takes months to reach users through a Chrome update, and other browsers are not guaranteed to honor it. Do not request preload unless you are sure you can serve HTTPS on the whole domain and every subdomain for the long term.","q":"How do I remove a domain from the HSTS preload list?"},{"a":"No. MDN says nosniff blocks script and style requests whose MIME type does not match the expected type. For other response types it only tells the browser to use the supplied Content-Type as-is instead of inferring one. It depends on your server setting Content-Type correctly.","q":"Does X-Content-Type-Options: nosniff stop all content-type confusion?"},{"a":"Not necessarily. MDN labels the header as Limited availability and Experimental, and says it is not Baseline because it does not work in some of the most widely used browsers. Check the browser compatibility table for the browsers your users run before relying on it.","q":"Is Permissions-Policy supported in every browser?"}],"markdown_url":"https://firewallsync.com/posts/http-security-headers-hsts-nosniff-referrer-policy-and-permissions-policy-explained/index.md","published":"2026-10-11","reading_minutes":9,"summary":"Four response headers do most of the work after Content Security Policy: Strict-Transport-Security, X-Content-Type-Options, Referrer-Policy and Permissions-Policy. Here is what each one does, the values MDN and OWASP document, and the HSTS preload trap that is hard to undo.","tags":["appsec"],"takeaways":["OWASP's recommended values are: Strict-Transport-Security: max-age=63072000; includeSubDomains; preload, X-Content-Type-Options: nosniff, Referrer-Policy: strict-origin-when-cross-origin, and Permissions-Policy: geolocation=(), camera=(), microphone=(). Treat them as starting points, not as drop-in answers for every site.","HSTS only takes effect after a browser has received the header once over HTTPS, and a browser ignores it if it arrives over plain HTTP. Preloading closes the first-visit gap but, per hstspreload.org, cannot easily be undone.","nosniff makes a browser block a script or stylesheet response whose Content-Type does not match, and stop guessing the type of other responses. It only works if your Content-Type headers are correct.","Modern browsers default to strict-origin-when-cross-origin for Referrer-Policy, but OWASP says to set it explicitly instead of relying on the default.","OWASP says not to set X-XSS-Protection (or to set it to 0) and not to use Expect-CT or Public-Key-Pins."],"title":"HTTP Security Headers Beyond CSP: HSTS, nosniff, Referrer-Policy and Permissions-Policy Explained","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-11","url":"https://firewallsync.com/posts/http-security-headers-hsts-nosniff-referrer-policy-and-permissions-policy-explained/"},{"author":"FirewallSync Editorial","description":"How to find overprivileged AWS IAM users and roles using Access Analyzer, CloudTrail activity, and policy comparison — before an incident finds them for you.","faq":[],"markdown_url":"https://firewallsync.com/posts/how-to-find-overprivileged-aws-iam-users-and-roles/index.md","published":"2026-10-10","reading_minutes":3,"summary":"Overprivileged IAM identities aren't usually the result of one bad decision. They're the accumulated residue of a hundred small, reasonable-seeming ones. Here's how to actually find them.","tags":["iam","cloud security"],"takeaways":[],"title":"How to Find Overprivileged AWS IAM Users and Roles","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-10-10","url":"https://firewallsync.com/posts/how-to-find-overprivileged-aws-iam-users-and-roles/"},{"author":"Tehseen Arbab","description":"Burp Suite vs OWASP ZAP compared: manual testing workflow versus automation-first design, licensing cost, and which fits a small team's actual testing needs.","faq":[],"markdown_url":"https://firewallsync.com/posts/burp-suite-vs-owasp-zap-which-web-app-scanner-fits-your-workflow/index.md","published":"2026-10-10","reading_minutes":4,"summary":"Both can find the same categories of web vulnerabilities. The real difference is workflow: one is built around a commercial tester's manual process, the other around free, scriptable automation.","tags":["tools"],"takeaways":[],"title":"Burp Suite vs OWASP ZAP: Which Web App Scanner Fits Your Workflow","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-10","url":"https://firewallsync.com/posts/burp-suite-vs-owasp-zap-which-web-app-scanner-fits-your-workflow/"},{"author":"FirewallSync Editorial","description":"How to write a strict Content Security Policy with nonces, what strict-dynamic and unsafe-inline do, and how to deploy with Report-Only and reporting endpoints.","faq":[{"a":"MDN describes CSP as a set of instructions from a site to the browser that restricts what the site's code is allowed to do. Its main use is defense in depth against cross-site scripting (XSS): even if an attacker manages to inject markup, a strict policy can stop the injected script from running. It does not remove the underlying injection bug.","q":"What does a Content Security Policy protect against?"},{"a":"MDN says nonces suit pages your server renders per request, and hashes suit static pages. A nonce is a random value generated for each response and placed in both the header and the script tags. A hash allows one specific script and breaks if the script changes at all, including whitespace, per OWASP.","q":"Should I use a nonce or a hash?"},{"a":"MDN advises avoiding it because it defeats much of the purpose of a CSP, since inline JavaScript is one of the most common XSS vectors. MDN also notes that when a directive contains nonce or hash expressions, browsers ignore 'unsafe-inline' in that directive.","q":"Is 'unsafe-inline' ever acceptable?"},{"a":"Yes. Send the policy in the Content-Security-Policy-Report-Only header. Violations are reported to your endpoint but nothing is blocked. If both headers are present, both are honored: the enforced one blocks and the report-only one generates reports.","q":"Can I test a policy without breaking my site?"},{"a":"MDN describes the frame-ancestors directive as a more flexible replacement for X-Frame-Options and recommends setting it to 'none' unless your site needs to be embeddable. We did not test every browser version; check your support requirements.","q":"Does CSP replace X-Frame-Options?"}],"markdown_url":"https://firewallsync.com/posts/content-security-policy-a-strict-nonce-based-policy-and-how-to-roll-it-out-safely/index.md","published":"2026-10-10","reading_minutes":6,"summary":"A Content Security Policy limits what scripts a browser will run, which can turn a cross-site scripting bug from a full compromise into a blocked request. Here is the strict nonce-based policy MDN and OWASP recommend, what each directive does, and a report-only rollout that does not break your site.","tags":["appsec","api-security"],"takeaways":["MDN's recommended practice for controlling script loading is a strict CSP: a policy built on per-response nonces or on hashes, not on lists of allowed hostnames.","A starting policy is script-src with a nonce, object-src 'none' and base-uri 'none'. Adding frame-ancestors 'none' blocks other sites from embedding yours unless you need that.","CSP is defense in depth. Both MDN and OWASP say you still have to sanitize input and encode output; OWASP states CSP should not be the only defense against cross-site scripting.","Deploy first with the Content-Security-Policy-Report-Only header. It reports violations but does not block anything, and it cannot be set through a meta element.","A nonce must be different for every HTTP response and unpredictable. A fixed nonce in a template is equivalent to no protection."],"title":"Content Security Policy: A Strict Nonce-Based Policy and How to Roll It Out Safely","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-10","url":"https://firewallsync.com/posts/content-security-policy-a-strict-nonce-based-policy-and-how-to-roll-it-out-safely/"},{"author":"FirewallSync Editorial","description":"What SPF, DKIM and DMARC each verify, how DMARC alignment works, the Gmail and Outlook.com bulk sender requirements, and the DMARC update in RFC 9989.","faq":[{"a":"To meet Google's bulk sender requirements (more than 5,000 messages per day to Gmail accounts) you need SPF, DKIM and a DMARC record, and the From domain must align with the SPF or DKIM domain. Microsoft's Outlook.com requirements for domains sending more than 5,000 emails per day also name all three. Below those volumes the requirements differ, but the three records together are what let a receiver tell your mail from a forgery.","q":"Do I need all three of SPF, DKIM and DMARC?"},{"a":"It is enough to satisfy the DMARC requirement in both the Google and Microsoft high-volume sender rules, which accept a policy of none. It does not ask receivers to quarantine or reject mail that fails DMARC, so it does not stop someone spoofing your domain. Its value is the reporting you receive while you find every legitimate sender.","q":"Is p=none enough?"},{"a":"RFC 7208 section 4.6.4 limits SPF evaluation to 10 DNS-querying terms: the include, a, mx, ptr and exists mechanisms and the redirect modifier. The ip4, ip6 and all mechanisms do not count. Going over the limit produces a permerror result, and DMARC then cannot use SPF as the passing method.","q":"What is the SPF 10-lookup limit?"},{"a":"SPF compares the connecting server's IP address with the list published by the envelope sender domain. A forwarding server is not on your list, so the check does not pass at the final recipient unless the forwarder rewrites the envelope sender. DKIM signatures usually survive simple forwarding because they travel inside the message, which is one reason to deploy both.","q":"Why does SPF often fail for forwarded mail?"},{"a":"RFC 9989 (May 2026) obsoletes RFC 7489 and changes some tags: pct is replaced by t, and np and psd are added. We did not test how individual receivers handle the new tags. Check each tag against your mail receivers' documentation before relying on a change.","q":"Does RFC 9989 replace my existing DMARC record?"}],"markdown_url":"https://firewallsync.com/posts/spf-dkim-and-dmarc-what-each-record-does-and-what-gmail-and-outlook-require/index.md","published":"2026-10-10","reading_minutes":10,"summary":"SPF, DKIM and DMARC answer three different questions about a message. Here is what each record checks, how alignment ties them together, what Google and Microsoft require from high-volume senders, and a rollout order that avoids blocking your own mail.","tags":["authentication","compliance"],"takeaways":["SPF checks whether the sending server's IP address is authorized for the envelope sender domain. DKIM checks a cryptographic signature added by a signing domain. DMARC requires one of them to pass and to align with the domain in the visible From address.","A DMARC record with policy p=none satisfies the Gmail and Outlook.com requirements for high-volume senders, but it asks receivers to take no action on spoofed mail. Protection starts at quarantine or reject.","Google's requirements have applied since February 1, 2024 to senders of more than 5,000 messages per day to Gmail accounts. Microsoft's Outlook.com requirements took effect May 5, 2025 for domains sending more than 5,000 emails per day.","SPF allows at most 10 DNS-querying terms per evaluation. Exceeding that returns a permanent error, which is why stacking vendor include entries eventually breaks the record.","In May 2026 the IETF published RFC 9989, a Standards Track version of DMARC that obsoletes RFC 7489. It replaces the pct tag with a t tag and adds np and psd tags."],"title":"SPF, DKIM and DMARC: What Each Record Does and What Gmail and Outlook Require","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-10","url":"https://firewallsync.com/posts/spf-dkim-and-dmarc-what-each-record-does-and-what-gmail-and-outlook-require/"},{"author":"Tehseen Arbab","description":"Open source vs commercial SIEM for small teams: what genuinely differs beyond license cost, and where each option's real hidden cost actually shows up.","faq":[],"markdown_url":"https://firewallsync.com/posts/open-source-vs-commercial-siem-what-actually-changes-at-small-scale/index.md","published":"2026-10-09","reading_minutes":5,"summary":"The license cost is the easiest number to compare and the least representative of the real cost. What actually differs at small scale is who's doing the engineering work the platform doesn't do for you.","tags":["tools"],"takeaways":[],"title":"Open Source vs Commercial SIEM: What Actually Changes at Small Scale","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-09","url":"https://firewallsync.com/posts/open-source-vs-commercial-siem-what-actually-changes-at-small-scale/"},{"author":"FirewallSync Editorial","description":"A practical guide to AWS IAM least privilege: how to find unused permissions with IAM Access Analyzer and safely tighten policies without breaking access.","faq":[],"markdown_url":"https://firewallsync.com/posts/aws-iam-least-privilege-a-practical-guide-to-finding-excess-permissions/index.md","published":"2026-10-09","reading_minutes":5,"summary":"Least privilege fails in practice because most teams have no visibility into which granted permissions are actually used. Here's how to find the gap and close it without breaking production.","tags":["iam","cloud security"],"takeaways":[],"title":"AWS IAM Least Privilege: A Practical Guide to Finding Excess Permissions","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-10-09","url":"https://firewallsync.com/posts/aws-iam-least-privilege-a-practical-guide-to-finding-excess-permissions/"},{"author":"FirewallSync Editorial","description":"Token theft explained: how adversary-in-the-middle attacks steal session cookies after MFA succeeds, and what detection actually needs to catch it.","faq":[],"markdown_url":"https://firewallsync.com/posts/what-is-token-theft-and-why-mfa-doesnt-always-stop-it/index.md","published":"2026-10-08","reading_minutes":4,"summary":"MFA verifies who's logging in, not who's still using the session afterward. Token theft attacks steal what MFA leaves behind: the authenticated session itself.","tags":["authentication","incident response"],"takeaways":[],"title":"What Is Token Theft and Why MFA Doesn't Always Stop It?","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-10-08","url":"https://firewallsync.com/posts/what-is-token-theft-and-why-mfa-doesnt-always-stop-it/"},{"author":"Tehseen Arbab","description":"1Password vs Bitwarden vs Vaultwarden compared for a small team: managed vs self-hosted tradeoffs, admin features, and which fits different team sizes and risk tolerances.","faq":[],"markdown_url":"https://firewallsync.com/posts/1password-vs-bitwarden-vs-vaultwarden-choosing-a-password-manager-for-a-small-team/index.md","published":"2026-10-08","reading_minutes":4,"summary":"The real decision isn't which one has more features. It's how much operational responsibility your team is willing to take on in exchange for lower cost and more control.","tags":["tools"],"takeaways":[],"title":"1Password vs Bitwarden vs Vaultwarden: Choosing a Password Manager for a Small Team","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-10-08","url":"https://firewallsync.com/posts/1password-vs-bitwarden-vs-vaultwarden-choosing-a-password-manager-for-a-small-team/"},{"author":"FirewallSync Editorial","description":"What the four S3 Block Public Access settings do, which one wins when they overlap, and how to enforce them across an AWS account or organization.","faq":[{"a":"Yes, for buckets created since AWS's April 2023 change. AWS said all newly created buckets get S3 Block Public Access enabled and ACLs disabled by default, whether created through the console, API, CLI, SDKs or CloudFormation. Buckets created before the change in your Region keep their existing settings.","q":"Are new S3 buckets private by default?"},{"a":"BlockPublicPolicy rejects attempts to attach a new bucket or access point policy that grants public access. RestrictPublicBuckets limits access to a bucket that already has a public policy, allowing only AWS service principals and authorized users in the bucket owner's account. You need both: one stops new mistakes, the other neutralizes old ones.","q":"What is the difference between BlockPublicPolicy and RestrictPublicBuckets?"},{"a":"No. AWS states that Block Public Access settings do not alter existing policies or ACLs. That is why removing a setting later makes a bucket with a public policy or ACL publicly accessible again.","q":"Does enabling Block Public Access change my bucket policies?"},{"a":"Keep the bucket private and serve it through Amazon CloudFront with origin access control, which grants the CloudFront distribution read access through a bucket policy. AWS recommends origin access control over the older origin access identity. If you must serve directly from S3, you have to relax the relevant settings for that one bucket.","q":"How do I host a static website if public access is blocked?"},{"a":"Yes. Since November 26, 2025, an AWS Organizations S3 policy can enable all four settings for the whole organization, an organizational unit or specific accounts, and new member accounts inherit it. It applies the four settings together; you cannot pick individual settings at the organization level.","q":"Can I enforce Block Public Access across all my AWS accounts?"}],"markdown_url":"https://firewallsync.com/posts/s3-block-public-access-what-the-four-settings-do-and-how-to-enforce-them/index.md","published":"2026-10-08","reading_minutes":9,"summary":"New S3 buckets have blocked public access by default since April 2023, but older buckets, account settings and new accounts can still be open. Here is what each setting does and how to enforce all four across an account or an AWS organization.","tags":["cloud-security","iam"],"takeaways":["S3 Block Public Access has four separate settings. Two stop new public ACLs and policies from being written; two make S3 ignore or restrict public ACLs and policies that already exist.","S3 applies the most restrictive combination of the organization, account, bucket and access point settings, so turning everything on at the account level covers every bucket in it.","Since April 2023, buckets created through the API, CLI, SDKs or CloudFormation get Block Public Access and ACLs disabled by default. Buckets created earlier keep whatever they had.","Since November 26, 2025, AWS Organizations can enforce all four settings across every account from one policy.","If you need a public website, put CloudFront in front of a private bucket with origin access control instead of making the bucket public."],"title":"S3 Block Public Access: What the Four Settings Do and How to Enforce Them","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-10-08","url":"https://firewallsync.com/posts/s3-block-public-access-what-the-four-settings-do-and-how-to-enforce-them/"},{"author":"FirewallSync Editorial","description":"CORS misconfiguration explained: what Access-Control-Allow-Origin with credentials allows, the origin-matching bugs attackers exploit, and a tested fix.","faq":[{"a":"It lets script on any website read the response, but only for requests made without credentials such as cookies. Browsers block access to the response if the request included credentials and the header is the wildcard. For public, non-personal data that is usually fine.","q":"What does Access-Control-Allow-Origin: * allow?"},{"a":"Not with Access-Control-Allow-Credentials: true. Reflecting every Origin is equivalent to allowing every website to read authenticated responses. Reflect the Origin only after it matches an exact entry in an allowlist.","q":"Is it safe to reflect the Origin header in Access-Control-Allow-Origin?"},{"a":"Browsers send Origin: null in several situations that an attacker can create, including requests from sandboxed iframes. PortSwigger's research describes using a sandboxed iframe to send a cross-origin request that carries Origin: null, which then matches an allowlist entry for null.","q":"Why is allowing the null origin dangerous?"},{"a":"No. CORS governs reading responses. A cross-site form post or a simple request is still sent with cookies, subject to SameSite rules, whatever the CORS headers say. Protect state-changing endpoints with CSRF defenses.","q":"Does CORS protect against CSRF?"},{"a":"No. CORS is enforced by browsers. Tools such as curl and server-side code ignore it, so authentication and authorization on the server are still required.","q":"Does CORS protect my API from non-browser clients?"}],"markdown_url":"https://firewallsync.com/posts/cors-misconfiguration-what-access-control-allow-origin-actually-permits/index.md","published":"2026-10-08","reading_minutes":7,"summary":"A CORS header decides which other websites may read your responses with your users' cookies attached. Reflecting the Origin header, matching it with endsWith or an unescaped regex, or trusting null each hands that read access to an attacker.","tags":["appsec","api-security"],"takeaways":["CORS does not block requests. It controls whether script on another website may read the response.","The dangerous combination is a specific Access-Control-Allow-Origin value plus Access-Control-Allow-Credentials: true, because that lets the other site read data fetched with your user's cookies.","Browsers refuse the wildcard * for credentialed requests, so the common shortcut is to echo back whatever Origin the request carried. That gives every website on the internet read access.","Suffix checks, prefix checks and regexes with unescaped dots accept attacker-registered domains. In our tests, a check for endsWith(\"example.com\") accepted https://evilexample.com.","Compare the Origin header against an exact list of allowed origins, never allow null, and send Vary: Origin."],"title":"CORS Misconfiguration: What Access-Control-Allow-Origin Actually Permits","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-08","url":"https://firewallsync.com/posts/cors-misconfiguration-what-access-control-allow-origin-actually-permits/"},{"author":"Tehseen Arbab","description":"What actually counts as PII, why classification is genuinely harder than checking a list of field names, and a practical approach to identifying it across real systems.","faq":[],"markdown_url":"https://firewallsync.com/posts/what-is-pii-and-why-classification-is-harder-than-it-sounds/index.md","published":"2026-10-07","reading_minutes":5,"summary":"PII sounds like it should be a simple list of field names to check for. In practice, whether something counts depends on context, combination with other data, and which regulation you're asking under.","tags":["privacy"],"takeaways":[],"title":"What Is PII, and Why Classification Is Harder Than It Sounds","topic":{"name":"Privacy \u0026 Third-Party Risk","slug":"privacy-risk","url":"https://firewallsync.com/categories/privacy-risk/"},"updated":"2026-10-07","url":"https://firewallsync.com/posts/what-is-pii-and-why-classification-is-harder-than-it-sounds/"},{"author":"FirewallSync Editorial","description":"How to secure OAuth 2.0 tokens in production: PKCE, refresh token rotation, sender-constrained tokens, and the mistakes RFC 9700 was written to stop.","faq":[],"markdown_url":"https://firewallsync.com/posts/how-to-secure-oauth-tokens-in-production/index.md","published":"2026-10-07","reading_minutes":5,"summary":"Most OAuth implementations follow the happy-path tutorial and skip the parts of the spec that exist specifically to stop token theft. Here's what the current best practice actually requires.","tags":["api security","authentication"],"takeaways":[],"title":"How to Secure OAuth Tokens in Production","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-07","url":"https://firewallsync.com/posts/how-to-secure-oauth-tokens-in-production/"},{"author":"FirewallSync Editorial","description":"CSRF protection compared: what SameSite=Lax leaves open, when you still need CSRF tokens, and how to reject cross-origin requests with Sec-Fetch-Site.","faq":[{"a":"Not on its own, according to OWASP's CSRF Prevention Cheat Sheet, which treats SameSite as defense in depth. Lax still sends cookies on cross-site top-level GET navigations, so any GET endpoint that changes state is exposed, and SameSite compares sites rather than origins, so a request from another subdomain of the same registrable domain is treated as same-site.","q":"Is SameSite=Lax enough to prevent CSRF?"},{"a":"OWASP's cheat sheet still recommends them: use your framework's built-in protection if it has one, otherwise add tokens to state-changing requests, and use Fetch Metadata checks with a fallback. A header check such as Sec-Fetch-Site can replace tokens for browsers that send it, which is the approach Go's standard library took in version 1.25, but you need a fallback for requests that arrive without the header.","q":"Do I still need CSRF tokens in 2026?"},{"a":"It is a request header that browsers add automatically to say how the page making a request relates to the server receiving it. The values are same-origin, same-site, cross-site and none, where none means a user-initiated action such as typing a URL. Page JavaScript cannot set or change it, because it is a forbidden request header.","q":"What is the Sec-Fetch-Site header?"},{"a":"If it authenticates with cookies, yes. If it authenticates only with a token that JavaScript places in an Authorization header, a cross-site page cannot make the browser attach that token, so classic CSRF does not apply. Many APIs accept both, and those need protection for the cookie path.","q":"Does a JSON API need CSRF protection?"},{"a":"In cross-site request forgery (CSRF) the attacker tricks a victim's browser into sending a request with the victim's cookies. In server-side request forgery (SSRF) the attacker tricks a server into sending a request from inside your network. The names are similar and the defenses are unrelated.","q":"What is the difference between CSRF and SSRF?"}],"markdown_url":"https://firewallsync.com/posts/csrf-protection-samesite-cookies-tokens-and-fetch-metadata-compared/index.md","published":"2026-10-07","reading_minutes":11,"summary":"SameSite cookies reduce cross-site request forgery but do not end it. Here is what each defense stops, where each one fails, and a header check you can add in a few lines.","tags":["appsec","authentication"],"takeaways":["SameSite=Lax stops cross-site POST requests from carrying your session cookie, but it works per site, not per origin, so a sibling subdomain is not blocked.","When a browser applies Lax as a default because no SameSite value was set, MDN documents a more permissive version: cross-site POSTs still carry cookies set in the previous two minutes. Set SameSite explicitly.","The Sec-Fetch-Site request header tells the server whether a request came from the same origin, the same site or another site. MDN lists it as available across browsers since March 2023.","Go 1.25, released in August 2025, added net/http.CrossOriginProtection, which blocks cross-origin state-changing requests using that header with no tokens.","OWASP's cheat sheet still lists CSRF tokens ahead of Fetch Metadata, and says a cross-site scripting flaw can bypass CSRF protections whichever you choose."],"title":"CSRF Protection: SameSite Cookies, Tokens and Fetch Metadata Compared","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-07","url":"https://firewallsync.com/posts/csrf-protection-samesite-cookies-tokens-and-fetch-metadata-compared/"},{"author":"FirewallSync Editorial","description":"CVSS vs EPSS vs KEV explained: what each score measures, where each fails, and a patch prioritization order tested against seven flaws exploited in late 2026.","faq":[{"a":"CVSS (Common Vulnerability Scoring System) rates how severe a vulnerability is on a 0 to 10 scale, based on characteristics such as how it is reached and what an attacker gains. EPSS (Exploit Prediction Scoring System) estimates the probability, from 0 to 1, that exploitation activity will be observed in the next 30 days. One describes potential damage, the other describes likelihood of attack.","q":"What is the difference between CVSS and EPSS?"},{"a":"Use both, in order. KEV is evidence that exploitation has happened, so a KEV listing outranks any EPSS score. EPSS is for the much larger set of vulnerabilities where no exploitation evidence exists yet. FIRST's EPSS guidance says that if direct evidence of active exploitation exists, that evidence should supersede EPSS.","q":"Should I use EPSS or KEV to prioritize patching?"},{"a":"EPSS is a statistical model built from observed exploitation data, and its FAQ says it generates daily scores without knowing whether a particular vulnerability was recently exploited. A newly disclosed flaw being used in targeted attacks may not yet have produced the kind of data the model sees. A low score soon after disclosure is not evidence of safety.","q":"Why does an exploited vulnerability have a low EPSS score?"},{"a":"No. The due dates in the catalog bind US Federal Civilian Executive Branch agencies, but the catalog is public and CISA says organizations should use it as an input to their vulnerability management prioritization framework.","q":"Is the CISA KEV catalog only for US government agencies?"},{"a":"Not by itself. A published CVSS base score describes the flaw in general, not your deployment. A 9.8 on a system that is not reachable from untrusted networks can reasonably wait behind a lower-scored flaw that is being exploited on an internet-facing system. CVSS itself provides Threat and Environmental metrics for adjusting a base score to your situation.","q":"Does a CVSS 9.8 mean I must patch immediately?"}],"markdown_url":"https://firewallsync.com/posts/cvss-vs-epss-vs-kev-which-signal-should-decide-what-you-patch-first/index.md","published":"2026-10-07","reading_minutes":12,"summary":"CVSS says how bad a flaw could be, EPSS estimates how likely it is to be attacked, and KEV records that it already has been. Seven recently exploited flaws show why the order you check them in matters.","tags":["vulnerability management","process"],"takeaways":["CVSS measures severity, EPSS estimates the probability of exploitation activity in the next 30 days, and CISA's KEV catalog lists flaws with reliable evidence of exploitation. They answer three different questions.","Check KEV first. FIRST, which publishes EPSS, says direct evidence of exploitation should supersede an EPSS score.","Seven flaws that CISA listed as exploited between September 27 and October 4, 2026 all had EPSS scores of 2.2% or lower on October 6, 2026. A 10% EPSS threshold would have flagged none of them.","A CVSS base score of 9.0 or higher would have flagged four of those seven. The other three scored 8.5 to 8.7.","None of the three knows your environment. Whether the affected system faces the internet is your input, and it changes the answer."],"title":"CVSS vs EPSS vs KEV: Which Signal Should Decide What You Patch First?","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-07","url":"https://firewallsync.com/posts/cvss-vs-epss-vs-kev-which-signal-should-decide-what-you-patch-first/"},{"author":"FirewallSync Editorial","description":"Passkeys vs passwords: how FIDO2's origin binding removes phishing as an attack path, and where recovery and enrollment gaps can undo that protection.","faq":[],"markdown_url":"https://firewallsync.com/posts/passkeys-vs-passwords-what-actually-changes-for-security-teams/index.md","published":"2026-10-06","reading_minutes":3,"summary":"Passkeys aren't just a more convenient password. They remove an entire attack category at the protocol level, but only if recovery and enrollment are governed correctly.","tags":["authentication","iam"],"takeaways":[],"title":"Passkeys vs Passwords: What Actually Changes for Security Teams?","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-10-06","url":"https://firewallsync.com/posts/passkeys-vs-passwords-what-actually-changes-for-security-teams/"},{"author":"Tehseen Arbab","description":"Data minimization explained: why collecting more than you need feels harmless upfront but creates compounding security and compliance costs, and how to actually practice it.","faq":[],"markdown_url":"https://firewallsync.com/posts/data-minimization-the-privacy-principle-most-teams-skip/index.md","published":"2026-10-06","reading_minutes":4,"summary":"Collecting more data than you need feels harmless because storage is cheap. The actual cost shows up later, at breach time, at deletion-request time, and at audit time — and by then it's a much bigger job.","tags":["privacy"],"takeaways":[],"title":"Data Minimization: The Privacy Principle Most Teams Skip","topic":{"name":"Privacy \u0026 Third-Party Risk","slug":"privacy-risk","url":"https://firewallsync.com/categories/privacy-risk/"},"updated":"2026-10-06","url":"https://firewallsync.com/posts/data-minimization-the-privacy-principle-most-teams-skip/"},{"author":"FirewallSync Editorial","description":"ZITADEL CVE-2026-105207, CVE-2026-105209 and CVE-2026-105215: affected 3.x and 4.x versions, the minimum fixed releases, workarounds, and how to check whether accounts were taken over.","faq":[{"a":"At least 4.17.3. That is the first release fixing CVE-2026-105207, and it is later than the fixes for the other two (4.17.1 and 4.16.2). ZITADEL's release notes for 4.19.2, dated September 28, 2026, say it addresses three further vulnerabilities (two rated High, one Low) and recommend that all 4.x deployments upgrade, so the better target is the newest 4.x release, which was 4.19.4 (October 1, 2026) when we checked. From 4.19.2, Login V2 requires a session cookie secret to be set before upgrading.","q":"Which ZITADEL version should I run to be protected from all three flaws?"},{"a":"Not completely. 3.4.15 fixes CVE-2026-105209 and CVE-2026-105215, but the advisory for CVE-2026-105207 lists 3.0.0 through 3.4.15 as affected and says the 3.x line has reached end-of-life, so the fix exists only in 4.x.","q":"I run ZITADEL 3.x. Am I covered if I update to 3.4.15?"},{"a":"Only partial ones. For CVE-2026-105207, disabling external identity provider account linking reduces the Login V2 route but not the API route. For CVE-2026-105215, disabling manual account creation on external identity providers helps, at the cost of self-service registration. ZITADEL's advisory for CVE-2026-105209 says no configuration workaround fully addresses it.","q":"Is there a workaround if I cannot upgrade today?"},{"a":"We found no report of exploitation. None of the three was in CISA's Known Exploited Vulnerabilities catalog (version 2026.10.04), and CISA's enrichment data on CVE-2026-105209 and CVE-2026-105215 lists exploitation as 'none' as of October 5, 2026. Absence of reports is not proof of absence, and details of all three are now public.","q":"Are these flaws being exploited?"},{"a":"CVSS (Common Vulnerability Scoring System) has more than one version. ZITADEL's advisories list 9.8, 9.6 and 9.1, which match the CVSS 3.1 scores in the CVE records. The same records carry CVSS 4.0 scores, and all three are 9.3.","q":"Why do the CVSS scores differ between sources?"}],"markdown_url":"https://firewallsync.com/posts/zitadel-account-takeover-cve-2026-105207-cve-2026-105209-cve-2026-105215-fixed-versions/index.md","published":"2026-10-06","reading_minutes":10,"summary":"Three critical ZITADEL flaws let attackers take over accounts through external identity provider linking, passkey enrollment and forged login callbacks. Fixes shipped between July and September 2026, but the CVE records only appeared on October 4, and ZITADEL 3.x has no fix for the worst one.","tags":["iam","authentication","vulnerability management"],"takeaways":["CVE-2026-105207 (account takeover through external identity provider linking) affects ZITADEL 4.0.0 through 4.17.2 and 3.0.0 through 3.4.15. The fix is 4.17.3 or later. ZITADEL's advisory says 3.x is end-of-life and has no fix, so 3.x users must move to 4.x.","CVE-2026-105209 (cross-organization takeover through passkey enrollment) is fixed in 4.17.1 and 3.4.15. It needs an attacker who already has user-write permission in some organization on the same instance. CVE-2026-105215 (account pre-hijacking in Login V1) is fixed in 4.16.2 and 3.4.14.","The fixes shipped before the CVE records: ZITADEL's releases page dates 3.4.15 and 4.17.1 to August 14, 2026 and 4.17.3 to September 4, 2026, while the CVE records were published on October 4, 2026.","As of October 5, 2026 none of the three was in CISA's Known Exploited Vulnerabilities catalog, and CISA's enrichment data lists exploitation as 'none' for two of them. That is a point-in-time statement, not a guarantee that no one has used them.","Some summaries show 9.8, 9.6 and 9.1 as the scores. Those match the older CVSS 3.1 scores. On CVSS 4.0, all three records score 9.3."],"title":"ZITADEL Account-Takeover Flaws CVE-2026-105207, -105209 and -105215: Fixed Versions and What to Check","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-10-06","url":"https://firewallsync.com/posts/zitadel-account-takeover-cve-2026-105207-cve-2026-105209-cve-2026-105215-fixed-versions/"},{"author":"FirewallSync Editorial","description":"CVE-2026-88779 in Citrix NetScaler ADC and Gateway: which SAML configurations are affected, the fixed builds (14.1-73.41, 13.1-64.28), what is confirmed about exploitation, and what to check.","faq":[{"a":"No. NetScaler's CVE record lists everything before 14.1-73.41 and before 13.1-64.28 as affected. Builds 14.1-73.37 and 13.1-64.23 are earlier than those, so they are in the affected range if SAML is configured.","q":"I upgraded to 14.1-73.37 or 13.1-64.23 after the September 27 bulletin. Am I protected against CVE-2026-88779?"},{"a":"No source we read confirms that. CISA's catalog entry describes a flaw that 'could allow for a denial of service', NetScaler's record describes denial of service, and watchTowr's published analysis focuses on denial of service only. SecurityWeek reported that watchTowr concluded the flaw can only be used to crash systems. Some reports describe attempts to download malware against honeypots, which shows attackers were probing, not that code execution through this flaw was achieved.","q":"Does CVE-2026-88779 allow remote code execution?"},{"a":"Only if you are a US Federal Civilian Executive Branch agency. CISA says Binding Operational Directive 26-04 applies to those agencies and encourages all organizations to adopt similar risk-based prioritization.","q":"Do I have to meet the CISA deadline of October 7, 2026?"},{"a":"Per NetScaler's description as relayed by watchTowr and The Hacker News, the issue requires the appliance to be configured as a SAML service provider or SAML identity provider. NetScaler's own CVE record lists affected versions without stating the condition, so confirm in your configuration and use the fixed builds regardless when you next upgrade.","q":"Is my appliance affected if it does not use SAML?"}],"markdown_url":"https://firewallsync.com/posts/netscaler-cve-2026-88779-saml-denial-of-service-exploited-patched-last-weeks-zero-days/index.md","published":"2026-10-06","reading_minutes":9,"summary":"CVE-2026-88779 can knock SAML-configured NetScaler ADC and Gateway appliances offline and is on CISA's exploited list. The fixed builds are newer than the ones that fixed last week's two zero-days, so an appliance patched on September 27 can still be vulnerable.","tags":["vulnerability management","incident response"],"takeaways":["CVE-2026-88779 is a memory overflow that can cause denial of service in NetScaler ADC and Gateway appliances configured as a SAML service provider or SAML identity provider. NetScaler's CVE record scores it 8.7 (High) on CVSS 4.0.","The fixed builds are 14.1-73.41, 13.1-64.28, 14.1-73.41 FIPS and 13.1-37.282. Builds from last week's bulletin (14.1-73.37 and 13.1-64.23) are older than these and fall inside the affected range.","CISA added the flaw to its Known Exploited Vulnerabilities catalog on October 4, 2026. The October 7 due date binds US federal civilian agencies only.","Confirmed impact is denial of service. Citrix says it has not identified an impact on customer data integrity, and watchTowr concluded the flaw can be used only to crash systems. Reports of remote code execution through this flaw are unconfirmed.","An appliance with no SAML authentication action or identity provider profile configured is outside the described condition, but check the configuration file rather than assuming."],"title":"NetScaler CVE-2026-88779: Exploited SAML Flaw Needs Newer Builds Than Last Week's Zero-Day Fixes","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-06","url":"https://firewallsync.com/posts/netscaler-cve-2026-88779-saml-denial-of-service-exploited-patched-last-weeks-zero-days/"},{"author":"FirewallSync Editorial","description":"CVE-2026-76504 in Cisco Catalyst SD-WAN Manager: which releases are fixed, what the Live Protect shield does and does not do, and the log entries Cisco says to review.","faq":[{"a":"Cisco's advisory lists these first fixed releases: 20.9.10.1 for the 20.9 train, 20.12.8.2 for 20.12, 20.15.6.1 for 20.15, 20.18.4.1 for 20.18, 26.1.2.1 for 26.1 and 26.2.1 for 26.2. It says releases earlier than 20.9 should migrate to a fixed release. Trains not named in the table have no fixed release listed, so ask Cisco TAC or check the advisory for updates.","q":"Which Cisco Catalyst SD-WAN Manager releases fix CVE-2026-76504?"},{"a":"No. Cisco's advisory says there are no workarounds that address the vulnerability. For on-premises deployments it recommends as a mitigation restricting access from unsecured networks such as the internet, and allowing only known trusted hosts if internet access is required. For Cisco-hosted cloud deployments Cisco says this mitigation is already deployed.","q":"Is there a workaround?"},{"a":"Cisco says it has addressed the flaw in Cisco SD-WAN Cloud (Cisco Managed) release 20.15.605 and that no customer action is required. You can check your remediation status and version through the Help function in the service GUI.","q":"Are Cisco-hosted SD-WAN cloud customers affected?"},{"a":"No. CISA's catalog entry sets October 3, 2026 under Binding Operational Directive 26-04, which applies to US Federal Civilian Executive Branch agencies. Other organizations are not bound by it, but the three-day window signals urgency.","q":"Does the CISA deadline apply to private companies?"}],"markdown_url":"https://firewallsync.com/posts/cisco-catalyst-sd-wan-manager-cve-2026-76504-authentication-bypass-fixed-releases-and-log-checks/index.md","published":"2026-10-05","reading_minutes":9,"summary":"Cisco says an unauthenticated authentication bypass in Catalyst SD-WAN Manager is being exploited. There is no workaround, only upgrades, a temporary shield and a short list of log entries worth checking today.","tags":["vulnerability management","incident response"],"takeaways":["Cisco advisory cisco-sa-sdwan-webauth-xr8beuuU, first published September 30, 2026, says an unauthenticated remote attacker can gain admin-level API access to Catalyst SD-WAN Manager by sending a specially encoded HTTP request. Cisco says it has seen active exploitation.","Every on-premises SD-WAN Manager is affected regardless of configuration. Cisco lists first fixed releases of 20.9.10.1, 20.12.8.2, 20.15.6.1, 20.18.4.1, 26.1.2.1 and 26.2.1. Cisco's table gives no fixed release for trains not listed in it, so confirm with Cisco if you run one.","Cisco lists no workaround. Its mitigation is to keep the manager off the internet. Its Live Protect shield, added to the advisory on October 2, 2026, gives only temporary partial protection and can block legitimate logins that use URI encoding.","Cisco's published indicators are entries mentioning j_security_check in two log files. Cisco's example uses %6a, but says any single encoded character can be used, so searching for that one string is not enough.","CISA added the flaw to its Known Exploited Vulnerabilities catalog on September 30, 2026 with an October 3 due date. That date binds US federal civilian agencies only."],"title":"Cisco Catalyst SD-WAN Manager CVE-2026-76504: Fixed Releases and Log Checks","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-05","url":"https://firewallsync.com/posts/cisco-catalyst-sd-wan-manager-cve-2026-76504-authentication-bypass-fixed-releases-and-log-checks/"},{"author":"FirewallSync Editorial","description":"CVE-2026-102489 and CVE-2026-102490 in the Zammad helpdesk: affected versions, what Zammad and DIVD each say, the unresolved root-escalation flaw, and what to do now.","faq":[{"a":"The CVE record and DIVD's case file list Zammad 6.3.0 through 6.5.4 as exploitable. DIVD also says the flaw is present in 7.0.0 through 7.1.3 but not exploitable because of environment conditions. Zammad's statement of October 1, 2026 says exploitation is only possible on 6.5 and older and that 7.0 and later are not affected. The CVE record's structured data stops at 'before 6.5.4', while its text says 'to 6.5.4', so treat 6.5.4 as affected.","q":"Which Zammad versions are affected by CVE-2026-102489?"},{"a":"We found none. DIVD's case file lists the patch status as available, but its recommendation is only to upgrade to version 7, and its affected range for this flaw runs up to 7.1.0-alpha. Zammad said on October 1, 2026 that it had received DIVD's details and was working on it, and pointed to its GitHub security advisories page for updates. We could not load that page from our environment, so check it directly.","q":"Is there a fix for CVE-2026-102490?"},{"a":"Zammad recommends it. Zammad says current versions are not affected by CVE-2026-102489 and that it hardened the affected code in 7.2.0, released September 23, 2026. For CVE-2026-102490, no source we read says 7.2.0 fixes it, so 7.2.0 should be treated as removing the remote entry point, not as a complete answer to the root escalation.","q":"Is Zammad 7.2.0 safe?"},{"a":"No. CISA set a due date of October 5, 2026 for both CVEs under Binding Operational Directive 26-04, which applies to US Federal Civilian Executive Branch agencies. Other organizations are not bound by it, but its forensic-triage requirement is a reasonable template for any exposed Zammad server.","q":"Does the CISA deadline apply to private companies?"}],"markdown_url":"https://firewallsync.com/posts/zammad-zero-days-cve-2026-102489-and-cve-2026-102490-what-is-fixed-and-what-is-not/index.md","published":"2026-10-05","reading_minutes":10,"summary":"Two Zammad flaws chained in the breach of the Dutch Institute for Vulnerability Disclosure are on CISA's exploited list. The sources disagree on which versions are safe, and no fix for the privilege-escalation flaw was visible as of October 5, 2026.","tags":["vulnerability management","incident response"],"takeaways":["CVE-2026-102489 (session hijack leading to code execution as the zammad user) affects Zammad 6.3.0 through 6.5.4 in exploitable form. Zammad says version 7.0 and later are not affected in practice, and that 6.5 and older no longer receive security fixes.","CVE-2026-102490 (local escalation from the zammad user to root) is listed by DIVD as affecting every version up to 7.1.0-alpha. As of October 5, 2026 we found no Zammad advisory or fixed version for it.","Zammad says CVE-2026-102490 cannot be exploited remotely on its own; an attacker needs prior access to the server. The 9.4 (Critical) score applies to the two flaws chained together, not to either one alone.","DIVD's advice to upgrade to version 7 closes the remote code execution path. By DIVD's own affected-version list it does not close the root escalation, so treat a patched server as still needing exposure limits and monitoring.","Both flaws are in CISA's Known Exploited Vulnerabilities catalog (added October 2, 2026). The October 5 due date binds US federal civilian agencies only."],"title":"Zammad Zero-Days CVE-2026-102489 and CVE-2026-102490: What Is Fixed and What Is Not","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-05","url":"https://firewallsync.com/posts/zammad-zero-days-cve-2026-102489-and-cve-2026-102490-what-is-fixed-and-what-is-not/"},{"author":"Tehseen Arbab","description":"What GDPR actually requires in practical terms for engineering teams: data minimization, the 72-hour breach rule, DPIAs, and what's genuinely an engineering decision versus a legal one.","faq":[],"markdown_url":"https://firewallsync.com/posts/what-gdpr-actually-requires-a-plain-english-breakdown-for-engineering-teams/index.md","published":"2026-10-05","reading_minutes":5,"summary":"GDPR compliance often gets treated as a legal problem handed to legal teams. Several of its core requirements are actually engineering decisions, and they're easier to satisfy if you know which ones fall on your desk.","tags":["privacy","compliance"],"takeaways":[],"title":"What GDPR Actually Requires: A Plain-English Breakdown for Engineering Teams","topic":{"name":"Privacy \u0026 Third-Party Risk","slug":"privacy-risk","url":"https://firewallsync.com/categories/privacy-risk/"},"updated":"2026-10-05","url":"https://firewallsync.com/posts/what-gdpr-actually-requires-a-plain-english-breakdown-for-engineering-teams/"},{"author":"FirewallSync Editorial","description":"How to detect leaked API keys across GitHub, logs, and client bundles, and rotate them safely without breaking the services that depend on them.","faq":[],"markdown_url":"https://firewallsync.com/posts/how-to-detect-and-rotate-leaked-api-keys-before-attackers-use-them/index.md","published":"2026-10-05","reading_minutes":4,"summary":"A leaked API key isn't automatically compromised, but you often can't prove it wasn't used. Here's how to find exposure fast and rotate without breaking production.","tags":["api security","secrets management"],"takeaways":[],"title":"How to Detect and Rotate Leaked API Keys Before Attackers Use Them","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-05","url":"https://firewallsync.com/posts/how-to-detect-and-rotate-leaked-api-keys-before-attackers-use-them/"},{"author":"FirewallSync Editorial","description":"Third-party SaaS risk assessment usually stops at a vendor questionnaire. A more useful vetting process focused on what actually predicts incidents.","faq":[],"markdown_url":"https://firewallsync.com/posts/vetting-third-party-saas-tools-before-they-touch-your-data/index.md","published":"2026-10-04","reading_minutes":2,"summary":"Most SaaS vetting stops at a security questionnaire the vendor's sales team fills out. Here's what actually predicts third-party risk in practice.","tags":["third-party risk","iam"],"takeaways":[],"title":"Vetting Third-Party SaaS Tools Before They Touch Your Data","topic":{"name":"Privacy \u0026 Third-Party Risk","slug":"privacy-risk","url":"https://firewallsync.com/categories/privacy-risk/"},"updated":"2026-10-04","url":"https://firewallsync.com/posts/vetting-third-party-saas-tools-before-they-touch-your-data/"},{"author":"Tehseen Arbab","description":"Encryption at rest vs encryption in transit: what each actually protects against, why compliance checklists treating them as one item miss a real gap, and where the boundary between them sits.","faq":[],"markdown_url":"https://firewallsync.com/posts/encryption-at-rest-vs-encryption-in-transit-what-each-actually-protects-against/index.md","published":"2026-10-04","reading_minutes":4,"summary":"Both terms get used as a single checkbox on a compliance form, but they protect against completely different threats — and having one without the other leaves a specific, predictable gap.","tags":["fundamentals","privacy"],"takeaways":[],"title":"Encryption at Rest vs Encryption in Transit: What Each Actually Protects Against","topic":{"name":"Fundamentals","slug":"fundamentals","url":"https://firewallsync.com/categories/fundamentals/"},"updated":"2026-10-04","url":"https://firewallsync.com/posts/encryption-at-rest-vs-encryption-in-transit-what-each-actually-protects-against/"},{"author":"FirewallSync Editorial","description":"Argon2id vs bcrypt vs scrypt: Argon2id at 19 MiB, 2 iterations, p=1 (OWASP) for new systems, bcrypt at cost 10+ (72-byte limit), and rehash on login.","faq":[{"a":"OWASP's Password Storage Cheat Sheet still lists bcrypt for legacy systems, with a work factor of 10 or more and a password limit of 72 bytes. It is not OWASP's first choice for new systems, which is Argon2id. An existing bcrypt deployment with a reasonable work factor does not need an emergency migration.","q":"Is bcrypt still safe to use in 2026?"},{"a":"OWASP's minimum is 19 MiB of memory, 2 iterations and 1 degree of parallelism, and it lists four other combinations it considers equivalent. RFC 9106 recommends higher settings: 64 MiB with 3 iterations and 4 lanes, or 2 GiB with 1 iteration and 4 lanes. Pick the highest setting your login servers can sustain at peak load.","q":"What Argon2id parameters should I use?"},{"a":"SHA-256 is designed to be fast, which is the wrong property for password storage. A salt stops an attacker from reusing one guess across many accounts, but each guess is still cheap. Password hashing functions such as Argon2id, scrypt and bcrypt are deliberately slow and let you raise the cost over time.","q":"Why not just use SHA-256 with a salt?"},{"a":"No. Argon2id, bcrypt and scrypt libraries generate a random salt for each password and store it inside the encoded hash string alongside the parameters. A pepper is different: it is a secret key shared across all passwords and must be stored separately from the database.","q":"Do I need to store the salt separately?"},{"a":"Verify the password against the old hash at login, and if it matches, hash the plaintext you now hold with Argon2id and replace the stored value. Accounts that never log in keep the old hash, so decide whether to expire those or wrap the old hashes inside the new algorithm.","q":"How do I move from bcrypt to Argon2id without resetting passwords?"}],"markdown_url":"https://firewallsync.com/posts/password-hashing-argon2id-vs-bcrypt-vs-scrypt-which-to-use-and-with-what-parameters/index.md","published":"2026-10-04","reading_minutes":13,"summary":"Use Argon2id for new systems, keep bcrypt if you already have it and handle its 72-byte limit, and use PBKDF2 only when FIPS requires it. The harder part is choosing parameters your login servers can afford.","tags":["authentication","appsec"],"takeaways":["For a new system, OWASP's first choice is Argon2id with at least 19 MiB of memory, 2 iterations and 1 degree of parallelism.","bcrypt is still acceptable for existing systems at a work factor of 10 or more, but most implementations only use the first 72 bytes of the password.","OWASP's numbers are minimums. RFC 9106 recommends more: 64 MiB with 3 iterations, or 2 GiB with 1 iteration, both with 4 lanes.","Memory cost is paid per login in progress. Multiply it by your peak concurrent logins before you raise it.","You can change algorithm or parameters without a password reset by rehashing each password the next time its owner logs in."],"title":"Argon2id vs bcrypt vs scrypt: OWASP Parameters to Use","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-04","url":"https://firewallsync.com/posts/password-hashing-argon2id-vs-bcrypt-vs-scrypt-which-to-use-and-with-what-parameters/"},{"author":"FirewallSync Editorial","description":"FortiMail CVE-2026-104286 (FG-IR-26-175) affects the 7.2, 7.4, 7.6 and 8.0 branches: affected builds, the disable-IBE workaround, log patterns and IPs to hunt.","faq":[{"a":"As of October 4, 2026, Fortinet's advisory FG-IR-26-175 tells customers to upgrade to upcoming FortiMail 8.0.2, 7.6.7 or 7.4.9 or above, and tells 7.2 customers to move to branch 7.4 or above. The word upcoming means those builds had not been released when the advisory text was written. Check the advisory for the current status before acting.","q":"Is there a patch for CVE-2026-104286?"},{"a":"Fortinet's advisory lists FortiMail 8.0.0 through 8.0.1, 7.6.0 through 7.6.6, 7.4.0 through 7.4.8 and 7.2.0 through 7.2.9. The structured version data in the CVE record differs, and also lists 7.0.0 through 7.0.9, so treat any build on those branches as needing a check against the advisory.","q":"Which FortiMail versions are affected?"},{"a":"Fortinet's advisory says to disable IBE feature support, either in the GUI under Encryption, IBE, by setting IBE Service to off, or from the CLI with config system encryption ibe, set status disable, end. As alternatives it lists removing internet access to the FortiMail webmail interface, or blocking POST requests to /ibe that contain ../ at a web application firewall.","q":"What is the workaround?"},{"a":"No. CISA's catalog entry sets a due date of October 4, 2026 under Binding Operational Directive 26-04, which applies to US Federal Civilian Executive Branch agencies. Other organizations are not bound by it, but the three-day window is a reasonable signal of urgency.","q":"Does the CISA deadline apply to private companies?"}],"markdown_url":"https://firewallsync.com/posts/fortimail-zero-day-cve-2026-104286-no-patch-yet-what-to-do/index.md","published":"2026-10-04","reading_minutes":12,"summary":"Fortinet says an unauthenticated file-write flaw in FortiMail is being exploited, and as of October 4, 2026 its advisory still lists the fixed builds as upcoming. The workaround is one setting. The compromise check takes longer.","tags":["vulnerability management","incident response"],"takeaways":["Fortinet's advisory FG-IR-26-175 lists FortiMail 8.0.0 through 8.0.1, 7.6.0 through 7.6.6, 7.4.0 through 7.4.8 and 7.2.0 through 7.2.9 as affected.","As of October 4, 2026 the advisory describes the fixed builds (8.0.2, 7.6.7 and 7.4.9) as upcoming. Until they ship, the fix is the workaround: turn off the IBE service.","The CVE record for the same flaw lists different affected and fixed versions than the advisory. Check your build against the advisory, not against a scanner result alone.","Fortinet published log patterns and two IP addresses seen in attacks, including a mail archive account that sends copies of mail to a remote server. Look for those before you assume the appliance is clean.","The CISA due date of October 4, 2026 binds US federal civilian agencies only."],"title":"FortiMail CVE-2026-104286: Affected Builds, IBE Workaround","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-04","url":"https://firewallsync.com/posts/fortimail-zero-day-cve-2026-104286-no-patch-yet-what-to-do/"},{"author":"Tehseen Arbab","description":"What a zero-day vulnerability actually is, how it differs from a regular unpatched CVE, and what a realistic, prioritized response looks like when one affects you.","faq":[],"markdown_url":"https://firewallsync.com/posts/what-is-a-zero-day-vulnerability-and-how-should-you-actually-respond-to-one/index.md","published":"2026-10-03","reading_minutes":4,"summary":"The term gets used loosely for anything urgent, but a zero-day specifically means a vulnerability being actively exploited before a fix exists — which changes what a reasonable response actually looks like.","tags":["fundamentals"],"takeaways":[],"title":"What Is a Zero-Day Vulnerability, and How Should You Actually Respond to One?","topic":{"name":"Fundamentals","slug":"fundamentals","url":"https://firewallsync.com/categories/fundamentals/"},"updated":"2026-10-03","url":"https://firewallsync.com/posts/what-is-a-zero-day-vulnerability-and-how-should-you-actually-respond-to-one/"},{"author":"FirewallSync Editorial","description":"Cloud misconfiguration is still the leading cause of cloud breaches. Static checklists miss drift — here's what to run instead.","faq":[],"markdown_url":"https://firewallsync.com/posts/cloud-misconfiguration-checklists-dont-work-heres-what-does/index.md","published":"2026-10-03","reading_minutes":2,"summary":"Checklist-based cloud security reviews catch what's on the list and miss everything that changed since it was written. Drift detection closes the gap a checklist can't.","tags":["cloud security","appsec"],"takeaways":[],"title":"Cloud Misconfiguration Checklists Don't Work — Here's What Does","topic":{"name":"Cloud Security","slug":"cloud-security","url":"https://firewallsync.com/categories/cloud-security/"},"updated":"2026-10-03","url":"https://firewallsync.com/posts/cloud-misconfiguration-checklists-dont-work-heres-what-does/"},{"author":"FirewallSync Editorial","description":"SSRF prevention for engineers: how server-side request forgery reaches cloud metadata, why deny lists fail, and how to layer validation, egress and IMDSv2.","faq":[{"a":"Server-side request forgery is a flaw where an application fetches a URL or network destination supplied by a user without ensuring the request goes where it should. The attacker uses the server as a relay to reach systems that only the server can reach, such as internal services or a cloud metadata endpoint.","q":"What is SSRF in simple terms?"},{"a":"It was its own category, A10, in the OWASP Top 10 2021. In the 2025 edition OWASP rolled SSRF into A01 Broken Access Control, so it is still covered but no longer listed separately.","q":"Is SSRF still in the OWASP Top 10?"},{"a":"No. IMDSv2 makes the AWS metadata service harder to reach through a typical SSRF bug because it requires a PUT request and a custom header before any metadata is returned. The SSRF flaw itself remains and can still be used against other internal services.","q":"Does IMDSv2 prevent SSRF?"},{"a":"Because the address your code checks is not always the address it connects to. A public URL can redirect to an internal one, a hostname can resolve differently on a second lookup, and the same address can be written in several notations. OWASP advises against relying on deny lists or regular expressions for these reasons.","q":"Why is a deny list of internal IP addresses not enough?"}],"markdown_url":"https://firewallsync.com/posts/ssrf-explained-how-server-side-request-forgery-works-and-how-to-prevent-it/index.md","published":"2026-10-03","reading_minutes":12,"summary":"Any feature that fetches a URL on a user's behalf lets that user borrow your server's network position. Here is how SSRF works, why URL filters keep failing, and the layered controls that hold.","tags":["appsec","api security","cloud security"],"takeaways":["SSRF happens when your server fetches a destination an attacker chose, so the request arrives with your server's network access and credentials rather than theirs.","The highest-value target in cloud environments is the instance metadata service at 169.254.169.254, which can return temporary credentials for the instance's role.","Checking the URL string is not enough: redirects, DNS answers that change between check and use, and alternate IP notations all get past string filters.","Use an allow list when destinations are known. When they are not, validate the resolved IP address, disable redirects and enforce the same rule at the network layer.","On AWS, requiring IMDSv2 removes the simplest metadata path, but AWS describes it as defense in depth, not a fix for the SSRF bug itself."],"title":"SSRF Explained: How Server-Side Request Forgery Works and How to Prevent It","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-03","url":"https://firewallsync.com/posts/ssrf-explained-how-server-side-request-forgery-works-and-how-to-prevent-it/"},{"author":"FirewallSync Editorial","description":"CVE-2026-88771 and CVE-2026-88772 in Citrix NetScaler ADC and Gateway: affected and fixed builds, exploitation timeline, hunting commands and what to rotate.","faq":[{"a":"According to the CVE records published by NetScaler, the fixed builds are NetScaler ADC and NetScaler Gateway 14.1-73.37 and 13.1-64.23, NetScaler ADC 14.1-73.37 FIPS, and NetScaler ADC 13.1-37.279 FIPS and NDcPP. Earlier builds on those branches are affected.","q":"Which NetScaler versions fix CVE-2026-88771 and CVE-2026-88772?"},{"a":"No vendor source says it does. Google's incident responders documented web shells, an altered Apache configuration and a SUID bit set on /bin/sh. Citrix's guidance for a suspected compromise is to preserve evidence, isolate the device, erase and reinstall firmware, restore a backup that predates the compromise and rotate secrets.","q":"Does upgrading remove an attacker who is already on the appliance?"},{"a":"Only if you are a US Federal Civilian Executive Branch agency. Binding Operational Directive 26-04 applies to those agencies. Private companies are not bound by it, although CISA encourages them to follow the same prioritization.","q":"Do I have to meet the CISA deadline of September 30, 2026?"},{"a":"No. Disabling DTLS or blocking inbound UDP port 443 upstream addresses the path used for CVE-2026-88772 only. CVE-2026-88771 is a separate command execution flaw that does not depend on DTLS, so the upgrade is still required.","q":"Is disabling DTLS enough?"}],"markdown_url":"https://firewallsync.com/posts/citrix-netscaler-zero-days-cve-2026-88771-cve-2026-88772-patch-and-hunt/index.md","published":"2026-10-03","reading_minutes":11,"summary":"Two NetScaler ADC and Gateway flaws were exploited before Citrix disclosed them on September 27, 2026. Upgrading closes the hole. It does not tell you whether someone already came through it.","tags":["vulnerability management","incident response"],"takeaways":["Fixed builds are NetScaler ADC and Gateway 14.1-73.37 and 13.1-64.23, plus 14.1-73.37 FIPS and 13.1-37.279 FIPS/NDcPP. Anything older on those branches is affected.","Both flaws were exploited before the September 27, 2026 bulletin, so an appliance that was internet-facing in September needs a compromise check, not only an upgrade.","CVE-2026-88772 is reached over DTLS on UDP port 443, which NetScaler documentation says is on by default for SSL VPN virtual servers.","Attackers were seen copying the appliance configuration directory, so treat stored credentials, certificates and private keys as exposed if you find signs of compromise.","The three-day CISA deadline binds US federal civilian agencies only, but it is a fair indicator of how urgent this is for everyone else."],"title":"Citrix NetScaler Zero-Days (CVE-2026-88771, CVE-2026-88772): Patch, Then Hunt","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-03","url":"https://firewallsync.com/posts/citrix-netscaler-zero-days-cve-2026-88771-cve-2026-88772-patch-and-hunt/"},{"author":"Tehseen Arbab","description":"Symmetric vs asymmetric encryption explained: how each actually works, why they're almost always used together in practice, and where each one fits.","faq":[],"markdown_url":"https://firewallsync.com/posts/symmetric-vs-asymmetric-encryption-when-to-use-each/index.md","published":"2026-10-02","reading_minutes":4,"summary":"Almost every real system uses both symmetric and asymmetric encryption together, each for the specific job it's actually good at. Understanding why explains a lot of how modern security protocols are built.","tags":["fundamentals"],"takeaways":[],"title":"Symmetric vs Asymmetric Encryption: When to Use Each","topic":{"name":"Fundamentals","slug":"fundamentals","url":"https://firewallsync.com/categories/fundamentals/"},"updated":"2026-10-02","url":"https://firewallsync.com/posts/symmetric-vs-asymmetric-encryption-when-to-use-each/"},{"author":"FirewallSync Editorial","description":"Security champions programs usually fail within two quarters for the same predictable reasons. What to structure differently from the start.","faq":[],"markdown_url":"https://firewallsync.com/posts/security-champions-programs-that-dont-fizzle-out/index.md","published":"2026-10-02","reading_minutes":2,"summary":"Most security champions programs launch with energy and quietly die within two quarters. The failure pattern is predictable, and so is the fix.","tags":["security culture","process"],"takeaways":[],"title":"Security Champions Programs That Don't Fizzle Out","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-10-02","url":"https://firewallsync.com/posts/security-champions-programs-that-dont-fizzle-out/"},{"author":"Tehseen Arbab","description":"The CIA triad explained: what confidentiality, integrity, and availability actually mean in practice, and why most real security decisions trade them off against each other.","faq":[],"markdown_url":"https://firewallsync.com/posts/what-is-the-cia-triad-and-why-it-still-shapes-security-decisions/index.md","published":"2026-10-01","reading_minutes":4,"summary":"Confidentiality, integrity, and availability are taught as a checklist. In practice they're better understood as three competing priorities that most security decisions are actually trading off against each other.","tags":["fundamentals"],"takeaways":[],"title":"What Is the CIA Triad, and Why It Still Shapes Security Decisions","topic":{"name":"Fundamentals","slug":"fundamentals","url":"https://firewallsync.com/categories/fundamentals/"},"updated":"2026-10-01","url":"https://firewallsync.com/posts/what-is-the-cia-triad-and-why-it-still-shapes-security-decisions/"},{"author":"FirewallSync Editorial","description":"Vulnerability management backlogs keep growing because most programs prioritize by CVSS score alone. A better way to triage what actually gets fixed.","faq":[],"markdown_url":"https://firewallsync.com/posts/the-vulnerability-backlog-that-never-shrinks/index.md","published":"2026-10-01","reading_minutes":2,"summary":"Vulnerability backlogs grow faster than teams can patch because most programs measure the wrong thing: total count instead of exploitability and exposure.","tags":["appsec","vulnerability management"],"takeaways":[],"title":"The Vulnerability Backlog That Never Shrinks","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-10-01","url":"https://firewallsync.com/posts/the-vulnerability-backlog-that-never-shrinks/"},{"author":"FirewallSync Editorial","description":"Security logging best practices for incident response differ from logging for dashboards. What to capture so you can actually reconstruct an incident.","faq":[],"markdown_url":"https://firewallsync.com/posts/logging-for-incidents-not-for-dashboards/index.md","published":"2026-09-30","reading_minutes":2,"summary":"Most logging strategies optimize for building dashboards, then fail the one test that matters: can you reconstruct what happened during an actual incident?","tags":["incident response","process"],"takeaways":[],"title":"Logging for Incidents, Not for Dashboards","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-09-30","url":"https://firewallsync.com/posts/logging-for-incidents-not-for-dashboards/"},{"author":"FirewallSync Editorial","description":"Software supply chain security doesn't require reviewing every dependency manually. A tiered audit approach that keeps release velocity intact.","faq":[],"markdown_url":"https://firewallsync.com/posts/dependency-audits-without-slowing-down-releases/index.md","published":"2026-09-29","reading_minutes":2,"summary":"Most supply-chain security advice assumes you can afford to review every dependency by hand. Here's a tiered approach that scales with a normal release cadence.","tags":["appsec","supply chain"],"takeaways":[],"title":"Dependency Audits Without Slowing Down Releases","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-09-29","url":"https://firewallsync.com/posts/dependency-audits-without-slowing-down-releases/"},{"author":"FirewallSync Editorial","description":"Container image scanning in CI catches CVEs but misses config drift and runtime secrets. Here's what to add to close the gap.","faq":[],"markdown_url":"https://firewallsync.com/posts/container-image-scanning-what-ci-pipelines-still-miss/index.md","published":"2026-09-28","reading_minutes":2,"summary":"Image scanning in CI catches known CVEs in base layers, but most pipelines still ship vulnerable configs and secrets that scanners aren't tuned to see.","tags":["appsec","cloud security"],"takeaways":[],"title":"Container Image Scanning: What CI Pipelines Still Miss","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-09-28","url":"https://firewallsync.com/posts/container-image-scanning-what-ci-pipelines-still-miss/"},{"author":"FirewallSync Editorial","description":"Most incident response tabletop exercises test whether people know the plan, not whether the plan survives contact with a real incident. Here's the fix.","faq":[],"markdown_url":"https://firewallsync.com/posts/the-tabletop-exercise-gap-testing-incident-response-for-real/index.md","published":"2026-09-27","reading_minutes":2,"summary":"Tabletop exercises satisfy the audit requirement but rarely test whether your team can actually execute under pressure. Here's the gap and how to close it.","tags":["incident response","process"],"takeaways":[],"title":"The Tabletop Exercise Gap: Testing Incident Response for Real","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-09-27","url":"https://firewallsync.com/posts/the-tabletop-exercise-gap-testing-incident-response-for-real/"},{"author":"FirewallSync Editorial","description":"Alert fatigue in security operations isn't a staffing problem — it's a signal your detection tuning has failed. Here's how to measure and fix it.","faq":[],"markdown_url":"https://firewallsync.com/posts/alert-fatigue-is-a-security-metric-not-a-morale-problem/index.md","published":"2026-09-26","reading_minutes":2,"summary":"Teams treat alert fatigue as a burnout issue to manage around. It's actually a leading indicator that your detection pipeline is broken.","tags":["incident response","process"],"takeaways":[],"title":"Alert Fatigue Is a Security Metric, Not a Morale Problem","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-09-26","url":"https://firewallsync.com/posts/alert-fatigue-is-a-security-metric-not-a-morale-problem/"},{"author":"FirewallSync Editorial","description":"MFA fatigue attacks exploit push notifications, not passwords. Here's why number matching and phishing-resistant MFA actually stop them.","faq":[],"markdown_url":"https://firewallsync.com/posts/mfa-fatigue-attacks-what-actually-stops-them/index.md","published":"2026-09-25","reading_minutes":2,"summary":"Push-notification bombing keeps working because most MFA rollouts treat every factor as equally trustworthy. Here's what actually closes the gap.","tags":["authentication","iam"],"takeaways":[],"title":"MFA Fatigue Attacks: What Actually Stops Them","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-09-25","url":"https://firewallsync.com/posts/mfa-fatigue-attacks-what-actually-stops-them/"},{"author":"FirewallSync Editorial","description":"Why leaked API keys keep showing up in breach reports despite better secrets scanning, and the three preventable mistakes behind most of them.","faq":[],"markdown_url":"https://firewallsync.com/posts/the-api-key-mistakes-that-keep-making-breach-reports/index.md","published":"2026-08-24","reading_minutes":2,"summary":"Secrets scanning tools have gotten good. The breaches keep happening anyway, mostly for three preventable reasons.","tags":["secrets management","appsec"],"takeaways":[],"title":"The API Key Mistakes That Keep Making Breach Reports","topic":{"name":"AppSec \u0026 APIs","slug":"appsec","url":"https://firewallsync.com/categories/appsec/"},"updated":"2026-08-24","url":"https://firewallsync.com/posts/the-api-key-mistakes-that-keep-making-breach-reports/"},{"author":"FirewallSync Editorial","description":"Why blameless postmortems often fail to prevent repeat incidents, and how to make action items stick so the same failure does not happen twice.","faq":[],"markdown_url":"https://firewallsync.com/posts/why-your-incident-postmortems-arent-preventing-repeats/index.md","published":"2026-08-12","reading_minutes":2,"summary":"Blameless postmortems became standard practice for good reasons, but most teams stopped halfway through adopting them — and it shows in the repeat incidents.","tags":["incident response","process"],"takeaways":[],"title":"Why Your Incident Postmortems Aren't Preventing Repeats","topic":{"name":"Security Operations","slug":"security-operations","url":"https://firewallsync.com/categories/security-operations/"},"updated":"2026-08-12","url":"https://firewallsync.com/posts/why-your-incident-postmortems-arent-preventing-repeats/"},{"author":"FirewallSync Editorial","description":"How to roll out least-privilege access without slowing engineers down: start from real usage data, phase the rollout, and make access requests fast enough that nobody routes around them.","faq":[],"markdown_url":"https://firewallsync.com/posts/least-privilege-access-controls-that-dont-slow-teams-down/index.md","published":"2026-08-01","reading_minutes":2,"summary":"Most least-privilege rollouts fail because they optimize for audit checklists instead of how engineers actually work. Here's a rollout order that holds up.","tags":["access control","iam"],"takeaways":[],"title":"Least-Privilege Access Controls That Don't Slow Teams Down","topic":{"name":"Identity \u0026 Access","slug":"identity-access","url":"https://firewallsync.com/categories/identity-access/"},"updated":"2026-08-01","url":"https://firewallsync.com/posts/least-privilege-access-controls-that-dont-slow-teams-down/"}],"description":"Practical cybersecurity writing for engineering teams that ship fast.","generated":"2026-11-21","license_and_citation":"https://firewallsync.com/for-ai/","llms_full_txt":"https://firewallsync.com/llms-full.txt","llms_txt":"https://firewallsync.com/llms.txt","site":"FirewallSync","url":"https://firewallsync.com/"}