Show Notes
1. GitLab path traversal at CVSS 10, now exploited — CVE-2026-85706 — Do: Upgrade GitLab, then grep for file.path — https://watchtowr.com/resources/rapid-reaction-gitlab-critical-path-traversal-vulnerability-cve-2026-85706/ 2. ScreenConnect file-transfer flaw now has a patch — CVE-2026-84869 — Do: Upgrade ScreenConnect to 26.6.5 — https://www.connectwise.com/company/trust/security-bulletins/2026-09-08-screenconnect-bulletin 3. Artifactory chain mints admins, drops a backdoor — CVE-2026-42018, CVE-2026-42016 — Do: Patch Artifactory, audit admin accounts — https://www.wiz.io/blog/artifactory-under-attack-in-the-wild-exploitation-of-cve-2026-42016-cve-2026-4201 4. Metasploit weaponizes five KEV flaws at once — CVE-2026-20079, CVE-2026-83549, CVE-2026-63077, CVE-2026-82078, CVE-2026-19295 — Do: Patch the five KEV flaws now weaponized — https://www.rapid7.com/blog/post/pt-metasploit-wrap-up-goes-to-sixteen 5. Plesk restore race gives a customer root — CVE-2026-68488 — Do: Upgrade Plesk to 18.0.80.7 — https://support.plesk.com/hc/en-us/articles/43248932867351-Vulnerability-in-Plesk-s-Backup-Manager-symlink-race-during-restore-allows-root-privilege-escalation 6. Dell ObjectScale unauth RCE at CVSS 10 — CVE-2026-70416 — Do: Upgrade ObjectScale to 4.4.0.0 — https://securityonline.info/dell-objectscale-vulnerabilities-cve-2026-70416/ 7. Direct Send phishing lands as internal mail — Do: Set RejectDirectSend to true in M365 — https://www.infosecurity-magazine.com/news/hackers-us-business-hours-m365/ 8. Malware hosted on AI vendors' own domains — Do: Filter AI share links as user content — https://www.huntress.com/blog/ai-attack-surface 9. Fake agency request passed every email check — Do: Verify agency data requests out of band — https://www.bleepingcomputer.com/news/security/revolut-discloses-data-breach-exposing-financial-info-passports/ 10. 1.8 million APKs mined for hardcoded secrets — Do: Scan your shipped mobile apps for secrets — https://www.bleepingcomputer.com/news/security/hackers-abused-claude-to-extract-secrets-from-18m-android-apps/ 11. Stolen police login opened a state DMV database — Do: Inventory external accounts into your systems — https://therecord.media/florida-shiny-hunters-motor-vehicle 12. Kiro wrote the edit before you approved it — CVE-2026-89332 — Do: Upgrade Kiro to 0.8.135, rotate creds — https://aws.amazon.com/security/security-bulletins/rss/2026-111-aws/ 13. 76% deployed Copilot, 43% reviewed permissions — Do: Review M365 oversharing before Copilot — https://www.infosecurity-magazine.com/news/organizations-skip-permissions-ai/ 14. IT staff clicked more than any other team — Do: Include IT in phishing simulations — https://www.securityweek.com/phishing-research-challenges-conventional-security-awareness-testing/ 15. OpenAI agent swarm ran the RubyGems attack — Do: Rotate RubyGems API keys from May — https://www.infosecurity-magazine.com/news/openai-agent-swarm-hacks-rubygems/
More from Daily DefSec Brief
Attackers are now exploiting PHP2Shell, a poisoned LiteLLM release carried a stealer for AI and cloud keys, and Fortinet shipped critical fixes while holding the CVEs until November 10. Every version number and every fix in this episode was checked against the vendor's own advisory or the researchers' report before it aired. Today's question: How do you check that your security alerts still reach a person? Reply in the comments or at [email protected].
PHP ships fixes today for two critical flaws whose details leaked ahead of disclosure, and a certificate authority's signing keys turned out to be guessable, one of them the number four. Also: a joint advisory on SafePay ransomware, a hijacked account that swapped OVHcloud's CLI downloads for a stealer, and a test account with the password Password1. Fixes were checked against the vendor's advisory or release notes where one exists, and against the researchers' report where not. Today's question: How does your team find accounts that should have been deleted? Reply in the comments or at [email protected].
The NetScaler flaw that crashed appliances is a step in a root takeover whose implant tries to copy itself to the HA partner. ClickFix crews bring a signed Siemens program to run their malware, Plex servers are being hijacked a day after an exploit went public, and a fake WatchGuard download site collects VPN logins. Every version number and every fix in this episode was checked against the vendor's advisory, release notes or the CVE record before it aired. Today's question: Where does your team tell people to get the VPN client? Reply in the comments or at [email protected].
Attackers used ConnectWise's app credentials to enter Microsoft 365 tenants. Also: a FortiGate implant, stolen UpdraftCentral keys, NTLMv1 blocks, AOMEI's update servers, and fixes. Every version number and every fix in this episode was checked against the vendor's advisory or the CVE record before it aired. Today's question: How do you review vendor apps in your tenant? Reply in the comments or at [email protected].
Attackers are using a BeyondTrust Remote Support zero-day and the Atlassian Data Center file-read flaw. Also: Office Preview Pane flaws, AI coding tokens stolen, Drupal, Next.js and NGINX fixes, hijacked NAS boxes, spyware in ads, and AI agents on leaked credentials. Every version number and every fix in this episode was checked against the vendor's own advisory before it aired. Today's question: Where do your developers' AI tool tokens live, and how long do they last? Reply in the comments or at [email protected].