Executive brief
MLflow, a platform for managing the machine learning lifecycle, contains a vulnerability where users with basic login access can view information about all registered machine learning models. This allows unauthorized users to see sensitive metadata, including model names, descriptions, and source locations, which could expose proprietary intellectual property in shared or multi-tenant environments. The issue is fixed in version 3.10.0.
Technical details
An improper access control vulnerability (CWE-284) exists in MLflow versions up to 3.9.0. The `SearchModelVersions` REST API endpoint and the `mlflowSearchModelVersions` GraphQL query fail to implement per-model authorization checks when basic authentication is enabled. This occurs because the endpoint was omitted from the `BEFORE_REQUEST_VALIDATORS` and `AFTER_REQUEST_HANDLERS` for REST, and from `GraphQLAuthorizationMiddleware.PROTECTED_FIELDS` for GraphQL. An authenticated attacker can exploit this to enumerate all model versions, tags, and source URIs across the entire registry. The vulnerability is resolved in version 3.10.0 by enforcing parent model permission filtering.
Affected products
- mlflow mlflow < 3.10.0
Timeline
- 2026-02-19: patched: Fix committed to repository
- 2026-05-21: disclosed: Vulnerability disclosed via huntr.dev and GHSA
- 2026-05-21: advisory: GitHub Advisory published