Info Disclosure in MLflow Tracking Server (CVE‑2026‑118624) Allows Unauthorized Data Access
What It Is — A vulnerability identified as CVE‑2026‑118624 affects the open‑source MLflow Tracking Server. The flaw resides in the _find_validator function, where an authorization check is omitted before the function returns job‑related data. An authenticated remote attacker can invoke the vulnerable endpoint and retrieve information that is scoped to the service account, resulting in a confidentiality impact (CVSS 5.4, Low confidentiality). The issue was reported on 2026‑06‑18, patched in MLflow 3.16.0, and publicly disclosed on 2026‑10‑29.
Affected Products — MLflow (all versions prior to 3.16.0).
Why It Matters for Trust & Control Assurance — The bug highlights a gap in the access‑control control objective: “Ensure that all privileged operations enforce proper authorization.” Verisq’s ACCESS_CONTROLS capability helps you evidence that such checks are in place and continuously monitored.
Recommended Actions —
- Upgrade all MLflow deployments to version 3.16.0 or later.
- Review any custom extensions or plugins for missing authorization checks.
- Enable detailed logging of job‑access requests and retain logs for audit purposes.
- Conduct a control‑assessment mapping of your authorization processes to the relevant VCF control objective.