CVE-2026-64849: MLflow Server-Side Request Forgery Vulnerability

## Overview CISA has added CVE-2026-64849 to its Known Exploited Vulnerabilities (KEV) catalog. This addition indicates a federal deadline for remediation. The vulnerability affects MLflow, an open-source AI engineering platform, allowing attackers to exploit server-side request forgery (SSRF) weaknesses. ## Technical Details The vulnerability resides in the POST /api/2.0/mlflow/webhooks/{id}/test endpoint. In versions prior to 3.15.0, this endpoint calls the _validate_webhook_url() function only for the original URL. However, the mlflow/webhooks/delivery.py component follows redirects and re-resolves hostnames without properly pinning the validated address. This flaw enables attackers to access internal or cloud metadata services, potentially exposing sensitive information. ## Impact Attackers can leverage this vulnerability to obtain response statuses and bodies from internal services. This could lead to unauthorized access to sensitive data, making it critical for organizations using MLflow to address this issue promptly. The CVSS score of 9.3 highlights the severity of the vulnerability. ## Mitigation Defenders should upgrade to MLflow version 3.15.0 or later to mitigate this vulnerability. Organizations should also review their configurations and access controls related to webhooks to minimize potential exploitation. Continuous monitoring for unusual activity is recommended to detect any attempts to exploit this vulnerability.