The MLflow platform suffers from a deserialization of untrusted data vulnerability in versions 1.23.0 and later, up to but excluding 2.14.2. This vulnerability enables attackers to upload a malicious LightGBM scikit-learn model that, when deserialized by the platform, can execute arbitrary code on the end user's system. This flaw arises from unsafe handling of serialized model data, allowing code execution upon interaction with the malicious model. The issue has been assigned CVE-2024-37056 and is rated high severity by the reporting database. A patch is available to remediate this vulnerability.

Successful exploitation allows arbitrary code execution on the system of an end user who interacts with a maliciously uploaded LightGBM scikit-learn model in the affected MLflow versions. This could lead to full compromise of the affected system depending on the privileges of the user running the MLflow platform.

A patch is available for this vulnerability. Users should upgrade MLflow to version 2.14.2 or later to remediate the issue. Until patched, avoid interacting with untrusted or unauthenticated model uploads, particularly LightGBM scikit-learn models. Review vendor advisories for official patch details and deployment guidance.