Vulnerability Overview
CVE-2026-0545 is a missing-authentication flaw in MLflow's FastAPI job routes when the Basic Auth application and job execution are enabled [S1]. MLflow's upstream authentication fix added protection and a regression test for these routes, and the fix is present in MLflow 3.11.1 and later supported releases [S2]. Red Hat Product Security independently rates the issue Important and documents affected product contexts [S3].
Impact
An unauthenticated network client may be able to interact with MLflow job-management functions despite the deployment's configured authentication boundary [S1]. The practical impact can include unauthorized access to job information, job disruption, resource consumption, and data exposure. Remote code execution is conditional: it depends on job execution being enabled and an allowed job function performing privileged work [S1][S3].
Remediation
Upgrade every running MLflow server to version 3.11.1 or newer and restart the deployment [S2]. Until the fixed runtime is live, set MLFLOW_SERVER_ENABLE_JOB_EXECUTION=false. Keep MLflow limited to trusted networks, VPN, SSO, or an authenticated reverse proxy where possible, and review access and job logs for unexpected activity [S3].
Covered by FixVibe
FixVibe verified active scans can confirm when exposed MLflow job routes fail to enforce the deployment's configured authentication boundary using bounded, read-only evidence. Findings identify the affected authentication boundary and explain what was verified. The scanner does not list, read, submit, execute, search, or cancel jobs, enumerate allowed job functions, or attempt remote code execution or denial of service.