CVE-2024-6838: MLflow Uncontrolled Resource Consumption vulnerability
MLflow 2.13.2 and earlier put no limit on the length of an experiment name. Creating or renaming an experiment with a very long string of digits makes the MLflow UI unresponsive. The `artifact_location` parameter also has no length cap. The advisory data lists no patched version.
Published Mar 20, 2025 · Updated Mar 22, 2025 · Analysis Nov 1, 2026
What an attacker can do
Anyone who can create or rename an experiment in the MLflow tracking server can give it an enormous name, for example a long run of integers. The UI panel that lists experiments becomes unresponsive for everyone who opens it. The advisory adds that artifact_location has no character limit at creation. The CVSS vector scores network access, no privileges and low availability impact.
MLflow tracking servers are often run without authentication inside a team network, so any user or script with access to the API can do this. The damage is a stuck dashboard and a service that is hard to use until the bad experiment is removed. The sources describe no data leak or code execution. The server may keep working for API clients while the browser UI hangs, but the sources do not test that.
Who is affected
The mlflow pip package at 2.13.2 and earlier. Any tracking server where untrusted users can reach the experiment API is in scope.
Check the version:
pip show mlflowThen list experiments through the UI or API and look for names that are unusually long.
mlflow
- Ecosystem
- PyPI
- Affected
- <= 2.13.2
- Fixed in
- No patched version yet
- Check
pip show mlflow
How to fix it
The advisory data lists no patched version and the sources give no fix commit. Upgrade to the latest MLflow release and read its notes for experiment name limits.
Limit exposure now. Put authentication in front of the tracking server, restrict who can create or rename experiments, and cap request size at a reverse proxy. If an oversized experiment is already breaking the UI, delete it with the MLflow API or CLI.
How to tell if you were hit
Check the experiment list for names that are very long or made of digits only, and for experiments with oversized artifact_location values. In access logs, look for experiments/create and experiments/update calls with unusually large request bodies.
Original advisory
From the GitHub Advisory Database (CC BY 4.0).
In mlflow/mlflow version v2.13.2, a vulnerability exists that allows the creation or renaming of an experiment with a large number of integers in its name due to the lack of a limit on the experiment name. This can cause the MLflow UI panel to become unresponsive, leading to a potential denial of service. Additionally, there is no character limit in the artifact_location parameter while creating the experiment.
Questions
- What breaks in CVE-2024-6838?
The MLflow UI panel becomes unresponsive when an experiment has a huge name. That is the denial of service the advisory describes.
- Does it need authentication?
The CVSS vector has no privileges required. MLflow has no authentication by default, so network reach to the API is what counts.
- Is there a patched MLflow version?
None is listed. Versions up to 2.13.2 are marked vulnerable. Use the latest release and restrict who can create experiments.
References
How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. The analysis is written by an AI model from the sources listed above and checked by automated rules before it is published. Check versions against the vendor's advisory before you act. Methodology
More MLflow vulnerabilities
All MLflow entriesCVE-2026-116690
MLflow vulnerable to arbitrary file read on unauthenticated servers via prompt model version source
high7.5Oct 27, 2026CVE-2026-116691
MLflow CrewAI autologging records LLM provider API keys in logged trace spans
medium6.5Oct 27, 2026CVE-2026-116692
MLflow vulnerable to remote code execution on unauthenticated servers via third-party scorer deserialization importing caller-placed artifact modules
critical9.8Oct 27, 2026CVE-2026-115886
MLflow: SSRF on unauthenticated MLflow servers via third-party scorer kwargs in the scorer invoke, register, and GenAI evaluate endpoints
medium6.5Oct 23, 2026CVE-2026-53995
MLflow FileStore trace IDs allow path traversal and arbitrary file writes
high8.2Oct 20, 2026GHSA-gqvg-gmmx-x4hm
MLFLOW_ALLOW_PICKLE_DESERIALIZATION=False safety control bypassed by mlflow.statsmodels flavor — RCE via crafted model artifact
high8.8Sep 1, 2026
