CVE-2026-69146

Received Received - Intake

Authenticated Log Input Injection in MLflow

Publication date: 2026-08-17

Last updated on: 2026-08-17

Assigner: GitHub, Inc.

Description

MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. From 3.13.0 until 3.15.0, LogInputs is absent from BEFORE_REQUEST_HANDLERS in the mlflow/server/auth package, allowing any authenticated user to call POST /api/2.0/mlflow/runs/log-inputs for another user's run_id and inject attacker-controlled DatasetInput records into the dataset_inputs lineage metadata without UPDATE permission. This issue is fixed in version 3.15.0.

CVSS Scores

EPSS Scores

Probability:
Percentile:

Meta Information

Published

2026-08-17

Last Modified

2026-08-17

Generated

2026-08-18

AI Q&A

2026-08-18

EPSS Evaluated

N/A

NVD

CVE-2026-69146

Affected Vendors & Products

Vendor	Product	Version / Range
mlflow	mlflow	3.15.0

Helpful Resources

Exploitability

CWE

KEV

CWE ID	Description
CWE-862	The product does not perform an authorization check when an actor attempts to access a resource or perform an action.

Attack-Flow Graph

Executive Summary

From versions 3.13.0 to 3.15.0, MLflow's authentication system lacked a LogInputs check in BEFORE_REQUEST_HANDLERS. This allowed authenticated users to inject malicious DatasetInput records into another user's run_id via the POST /api/2.0/mlflow/runs/log-inputs endpoint without needing UPDATE permission.

Detection Guidance

Detecting this vulnerability requires checking if your MLflow server version is between 3.13.0 and 3.15.0 and verifying if unauthorized POST /api/2.0/mlflow/runs/log-inputs requests are being made. Inspect server logs for suspicious log-inputs API calls targeting other users' run IDs. Ensure no attacker-controlled DatasetInput records appear in lineage metadata without proper permissions.

Impact Analysis

An attacker could manipulate lineage metadata in MLflow, potentially altering data provenance or injecting false records. This could lead to incorrect model training data, compromised audit trails, or misleading research outcomes if undetected.

Compliance Impact

This vulnerability allows unauthorized users to inject malicious DatasetInput records into lineage metadata without proper permissions. This could lead to data integrity issues, potentially violating GDPR's data accuracy requirements or HIPAA's integrity principles by allowing unauthorized modifications to sensitive data.

Mitigation Strategies

Upgrade MLflow to version 3.15.0 or later to address the missing LogInputs protection in the auth package.

Hi! I’m here to help you understand CVE-2026-69146. Ask me anything about the vulnerability, its impact, or mitigation strategies.

0/70