MLflow: Remote code execution on unauthenticated MLflow servers via third-party scorer deserialization importing caller-placed artifact modules
- Published
- Last modified
- Status
- published
- Known exploitation
- Not listed in CISA KEV
- Fix
- Fixed in 3.17.0
- Public exploit
- None on Exploit-DB
Summary
MLflow, an open source platform for building and managing AI agents, large language models, and machine learning models, is affected in versions from 3.12.0rc0 up to (but not including) 3.17.0. A remote attacker who can reach an MLflow server with authentication disabled, or who holds ordinary experiment and artifact-upload permissions under basic authentication, could run arbitrary code on the server by planting a malicious Python module where the scorer loader will import it.
Summarized by anthropic/claude-haiku-5.5. Verify details against the original sources.
Read the original description
MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. From 3.12.0rc0 until 3.17.0, third-party scorer deserialization in mlflow/genai/scorers/base.py accepts request-selected descendants of THIRD_PARTY_SCORER_ALLOWED_MODULES rather than requiring an exact allowed module. A caller can use MLflow artifact endpoints to place a Python module beneath a writable allowed package and submit its module and class through /ajax-api/3.0/mlflow/scorer/invoke. Scorer.model_validate_json() imports the module and constructs the class before trace processing. Authentication-disabled deployments permit this code execution without credentials, while basic-auth deployments require a principal capable of creating an experiment and run and uploading artifacts. In either configuration, the allowed scorer package must be writable by the server process. This issue is fixed in version 3.17.0.
Affected versions
- mlflow mlflow
From 3.12.0rc0 · fixed in 3.17.0
- mlflowPyPI
From 3.12.0rc0 · fixed in 3.17.0