PT-2026-76821 · Mlflow · Mlflow
CVE-2026-64849
·
Published
2026-07-02
·
Updated
2026-09-02
CVSS v3.1
9.3
Critical
| Vector | AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:L/A:N |
Name of the Vulnerable Software and Affected Versions
MLflow versions prior to 3.15.0
Description
An unauthenticated Server-Side Request Forgery (SSRF) exists in the MLflow Tracking Server. The
POST /api/2.0/mlflow/webhooks/{id}/test endpoint allows an attacker to make the server issue HTTP requests to arbitrary internal, loopback, or cloud-metadata endpoints and read the responses. While a validation function validate webhook url() is used to reject non-public IPs, it only checks the original URL and does not pin the validated address. An attacker can bypass this guard by hosting a public HTTPS endpoint that returns a 302 redirect to an internal target, which the server follows without re-validation. This can be used to steal cloud instance-metadata (such as AWS IMDS IAM credentials), access internal admin services, or perform internal port scanning. Additionally, 307 or 308 redirects can be used for blind write attacks against private-network management endpoints.Recommendations
Update MLflow to version 3.15.0 or later.
As a temporary workaround, restrict network access to the MLflow Tracking Server to trusted sources only to prevent unauthenticated access to the webhooks API.
Exploit
Fix
SSRF
Found an issue in the description? Have something to add? Feel free to write us 👾
Weakness Enumeration
Related Identifiers
Affected Products
Mlflow