The vulnerability CVE-2025-15381 in mlflow/mlflow arises because tracing and assessment endpoints lack proper permission validation when the basic-auth app is enabled. Consequently, authenticated users with no permissions on experiments can read trace information and create assessments for traces they should not access. This leads to exposure of sensitive trace metadata (confidentiality impact) and unauthorized modification via assessment creation (integrity impact). The vulnerability is present in deployments running 'mlflow server --app-name=basic-auth'.

The vulnerability impacts confidentiality by exposing trace metadata to unauthorized authenticated users. It also impacts integrity by allowing these users to create assessments for traces without proper permissions. There is no indication of availability impact. The CVSS 3.0 score is 8.1 (high severity), reflecting network attack vector, low attack complexity, low privileges required, no user interaction, and high confidentiality and integrity impacts.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. The Red Hat advisory linked does not explicitly state a fix or workaround. Until a patch or official fix is available, deployments should consider disabling the basic-auth app or restricting authenticated user access to trusted users only. Monitor the vendor advisory for updates on patches or official mitigations.