NewApp Monitoring is liveSign up free
AgentbeamBeta
AI agent vulnerability · MLflow

GHSA-83fm-w79m-64r5: Remote file access vulnerability in `mlflow server` and `mlflow ui` CLIs

critical

MLflow servers started with mlflow server or mlflow ui, on versions before 2.3.1, allow remote file access through the Model Registry for anyone who can send requests to them. Fixed in MLflow 2.3.1, released to PyPI on April 27th, 2023. Servers that anyone can reach should be upgraded and access-restricted.

Published May 1, 2023 · Analysis Nov 2, 2026

What an attacker can do

The maintainers describe a remote file access exploit that is very similar to CVE-2023-1177, which is also in our database. It applies to the Model Registry served by the mlflow server and mlflow ui commands. The advisory gives no endpoint or payload, so none is listed here.

The condition is that nobody limits who can query the server. Stock MLflow has no authentication, so any client with network access can attempt it, and the attacker gets files the server can reach: local files, and remote files held in cloud blob storage when the server holds broad cloud credentials.

Products that do not run these commands are not affected. The advisory names Databricks Managed MLflow and MLflow on Azure Machine Learning as unaffected. This advisory has no CVSS data in our record, but it is rated critical.

Who is affected

pip package mlflow older than 2.3.1, when hosted with mlflow server or mlflow ui. Check:

pip show mlflow

Integrations that use MLflow as a client library only, without running those commands, are not affected.

mlflow

Ecosystem
PyPI
Affected
< 2.3.1
Fixed in
2.3.1
Check
pip show mlflow

How to fix it

Upgrade to MLflow 2.3.1 or later:

pip install --upgrade "mlflow>=2.3.1"

The advisory also asks you to restrict access with a cloud VPC, an IP allowlist for inbound requests, or authentication and authorization middleware. Limit the remote files the server can reach too. If artifacts live in cloud blob storage, scope the server's cloud credentials to only the files and directories MLflow needs.

How to tell if you were hit

Review access logs of the MLflow server for Model Registry requests from unknown clients and for source values or paths that point outside your artifact root. Check cloud storage audit logs for reads by the MLflow server's identity of buckets or prefixes unrelated to MLflow. Check which hosts can reach the port.

Original advisory

From the GitHub Advisory Database (CC BY 4.0).

Impact

Users of the MLflow Open Source Project who are hosting the MLflow Model Registry using the `mlflow server or mlflow ui` commands using an MLflow version older than MLflow 2.3.1 may be vulnerable to a remote file access exploit if they are not limiting who can query their server (for example, by using a cloud VPC, an IP allowlist for inbound requests, or authentication / authorization middleware).

This issue only affects users and integrations that run the `mlflow server and mlflow ui commands. Integrations that do not make use of mlflow server or mlflow ui` are unaffected; for example, the Databricks Managed MLflow product and MLflow on Azure Machine Learning do not make use of these commands and are not impacted by these vulnerabilities in any way.

The vulnerability is very similar to https://nvd.nist.gov/vuln/detail/CVE-2023-1177, and a separate CVE will be published and updated here shortly.

Patches

This vulnerability has been patched in MLflow 2.3.1, which was released to PyPI on April 27th, 2023. If you are using `mlflow server or mlflow ui` with the MLflow Model Registry, we recommend upgrading to MLflow 2.3.1 as soon as possible.

Workarounds

If you are using the MLflow open source `mlflow server or mlflow ui` commands, we strongly recommend limiting who can access your MLflow Model Registry and MLflow Tracking servers using a cloud VPC, an IP allowlist for inbound requests, authentication / authorization middleware, or another access restriction mechanism of your choosing.

If you are using the MLflow open source `mlflow server or mlflow ui` commands, we also strongly recommend limiting the remote files to which your MLflow Model Registry and MLflow Tracking servers have access. For example, if your MLflow Model Registry or MLflow Tracking server uses cloud-hosted blob storage for MLflow artifacts, make sure to restrict the scope of your server's cloud credentials such that it can only access files and directories related to MLflow.

References

Questions

Is this the same as CVE-2023-1177?

It is described as very similar to CVE-2023-1177, with a separate CVE to follow. Both are in the database, and 2.3.1 is the fixed release here.

Is Databricks Managed MLflow affected?

No. The advisory says Databricks Managed MLflow and MLflow on Azure Machine Learning do not use these commands and are unaffected.

What if I cannot upgrade right away?

Restrict who can query the server with a VPC, IP allowlist or auth middleware, and narrow the cloud credentials the server holds.

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 entries