Missing Authorization in MLflow Experiment-scoped Label Schema CRUD API

Executive Summary

This vulnerability in MLflow (CVE-2026-13484) involves the Experiment-scoped Label Schema CRUD API bypassing Role-Based Access Control (RBAC) checks. Specifically, new label schema APIs were introduced but their protobuf request classes were not registered properly in the authentication handlers. As a result, authenticated non-admin users can perform create, read, update, and delete operations on label schemas without having the required experiment permissions.

The vulnerability allows these unauthorized operations after basic authentication succeeds, effectively missing authorization checks. The issue affects multiple MLflow components such as tracking, evaluation, and UI/UX.

Useful 0 Not Useful 0

Impact Analysis

This vulnerability can impact you by allowing authenticated users with limited permissions to bypass experiment-level access controls and manipulate label schemas. This means unauthorized users could create, read, update, or delete label schemas related to experiments without proper authorization.

Such unauthorized access could lead to data integrity issues, unauthorized changes in experiment metadata, and potential disruption of AI model tracking and evaluation processes.

Useful 0 Not Useful 0

Detection Guidance

This vulnerability involves unauthorized access to the experiment-scoped label schema CRUD APIs in MLflow, specifically under the endpoints `/api/3.0/mlflow/label-schemas/*` and `/ajax-api/3.0/mlflow/label-schemas/*`.

To detect exploitation attempts on your system or network, you can monitor HTTP requests targeting these label schema API endpoints, especially those made by authenticated non-admin users.

Suggested commands include using network monitoring or web server logs to filter requests to these paths. For example, using `grep` on access logs:

• grep "/api/3.0/mlflow/label-schemas/" /path/to/mlflow/access.log
• grep "/ajax-api/3.0/mlflow/label-schemas/" /path/to/mlflow/access.log

Additionally, you can check for unexpected create, read, update, or delete operations on label schemas by reviewing logs or audit trails for these API calls.

Useful 0 Not Useful 0

Mitigation Strategies

Immediate mitigation involves restricting access to the vulnerable label schema APIs and ensuring proper authorization checks are enforced.

Specifically, you should:

• Limit access to the endpoints `/api/3.0/mlflow/label-schemas/*` and `/ajax-api/3.0/mlflow/label-schemas/*` to trusted administrators only.
• Apply network-level controls such as firewall rules or API gateway policies to block unauthorized users from accessing these endpoints.
• Monitor and audit API usage to detect any unauthorized operations.

Longer term, apply the recommended fix by updating MLflow to include the label schema protobuf request classes in the authentication handlers and implement the necessary experiment and schema-id based permission validators as described in the fix for this issue.

Useful 0 Not Useful 0

Compliance Impact

The vulnerability in MLflow allows authenticated non-admin users to bypass experiment Role-Based Access Control (RBAC) checks on label schema APIs, enabling unauthorized create, read, update, and delete operations on label schemas.

Such missing authorization could lead to unauthorized access or modification of sensitive data or metadata related to experiments, which may impact compliance with data protection regulations like GDPR or HIPAA that require strict access controls and data integrity.

However, the provided information does not explicitly state the direct impact on compliance with these standards or regulations.

Useful 0 Not Useful 0