CVE-2026-1168 is a denial-of-service flaw in GitLab CE/EE caused by improper resource allocation limits in the GraphQL complexity calculation logic (CWE-770). An unauthenticated attacker can send crafted GraphQL queries to a GitLab instance and, under certain conditions, exhaust server resources and render the instance unavailable. The attacker gains no code execution or data access — the impact is availability only, but the CVSS 3.1 score of 7.5 (AV:N/AC:L/PR:N) reflects that exploitation requires only network access with no privileges or user interaction. All self-managed GitLab CE and EE deployments in the affected version ranges are exposed, particularly those with the GraphQL API reachable from the internet. No public proof-of-concept, CISA KEV listing, or confirmed in-the-wild exploitation is known; remediation is available in patched releases.
What to do: Upgrade self-managed GitLab CE/EE to 19.1.8, 19.2.6, or 19.3.2 (or later), depending on your current release branch. As an interim mitigation, restrict the GraphQL endpoint (/api/graphql) to trusted networks or authenticated clients at the reverse proxy and apply rate limiting to blunt resource-exhaustion attempts. Monitor CPU/memory usage and instance availability for signs of abuse, and check the GitLab version shown in /api/v4/version or /help to confirm you are out of the affected ranges.
| Product | Affected |
|---|
| GitLab CE (Community Edition) | >= 18.4.6 and < 19.1.8; >= 19.2 and < 19.2.6; >= 19.3 and < 19.3.2 |
| GitLab EE (Enterprise Edition) | >= 18.4.6 and < 19.1.8; >= 19.2 and < 19.2.6; >= 19.3 and < 19.3.2 |
Estimated exposure
largehundreds of thousands of self-managed GitLab CE/EE installations, with public internet scans consistently showing tens of thousands of exposed GitLab instances — GitLab is one of the most widely deployed source-code platforms (tens of millions of users), and independent internet scans (Shodan/Censys-style surveys) repeatedly enumerate on the order of 50,000–100,000 internet-facing GitLab instances,…
Order-of-magnitude estimate by the model from install counts, market share and public scan data it knows; verify before quoting.