Posted in Security Operations
"We patched it" and "we're not vulnerable" are two different statements. Alpine Linux just showed why…
Craig M. — Chief Product Officer @ Cyberr®
"We patched it" and "we're not vulnerable" are two different statements. Alpine Linux just showed why.
CVE-2026-97229 is a CVSS 10 GitLab flaw: a crafted regex in a CI config can give an unauthenticated attacker code execution. CISA added it to KEV last week, and Shadowserver still counted over 1,000 exposed servers on 2 November.
Alpine upgraded GitLab after the fix shipped. Their instance was still vulnerable.
The reason: the fix is a patch to Ruby, which GitLab applies only in its official packages and images. Alpine ran GitLab on standard Ruby. The GitLab version number said fixed. The runtime underneath said otherwise. Meanwhile a freshly registered account crashed their workers eight times, and with no outbound connection logging they couldn't rule out compromise. So they rebuilt from scratch.
The gap most teams miss: scanners and patch trackers key off the application version. If you build your own images or swap in a distro runtime, the advisory may not describe your system.
What I'd check this week:
→ Inventory self-managed GitLab, including custom images, and confirm the Ruby version (3.3.13, 3.4.12, 4.0.9 or later), not just GitLab's.
→ Restrict open sign-ups and who can run pipelines on internet-facing instances.
→ If you were exposed, rotate runner tokens, CI/CD variables and personal access tokens, not only passwords.
→ Log outbound connections from build infrastructure. Without them, "no evidence" just means no data.
How does your vulnerability management handle fixes that ship in a dependency rather than the product itself?
#VulnerabilityManagement #DevSecOps #CICD
6 reactions · 0 comments
Topics: #devsecops, #vulnerabilitymanagement, #cicd