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

Should you act?

Exploited: patch now if you run it

CISA has confirmed exploitation in the wild. This signal is computed from CISA KEV, EPSS and CVSS data; it is not an editorial verdict.

• Yes, if you run MLflow MLflow. CISA added this CVE to its Known Exploited Vulnerabilities catalog on 2026-08-19, which means there is reliable evidence of exploitation in the wild. US federal agencies must remediate by 2026-09-02; treat that as the outer limit, not the target.
• EPSS 9.8%, percentile 96: FIRST’s model estimates a 9.8% chance of exploitation activity in the next 30 days, higher than 96% of all scored CVEs (very high). EPSS is a probability, not a verdict.
• Remotely reachable with no login and no user interaction. If the vulnerable service is exposed to the internet, prioritize it.
• Only applies if your dependency tree includes mlflow (PyPI) >= 3.3.0, < 3.15.0, fixed in 3.15.0. Check your lockfile, including transitive dependencies.
• Editorial verdict: Act now. MLflow CVE-2026-64849: Exploited Unauthenticated SSRF — If an MLflow tracking server below 3.15.0 is reachable by anyone you do not fully trust, assume attackers can read your cloud instance credentials through it: upgrade today, then check webhooks and cloud audit logs.

What it is

MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. Starting in 3.3.0 and prior to 3.15.0, the unauthenticated POST /api/2.0/mlflow/webhooks/{id}/test endpoint calls validatewebhookurl() in mlflow/utils/validation.py only for the original URL while mlflow/webhooks/delivery.py follows redirects and re-resolves the hostname without pinning the validated address, allowing attackers to reach internal or cloud metadata services and receive responsestatus and responsebody. This issue is fixed in version 3.15.0.

From the GitHub advisory

Summary The default MLflow Tracking Server (mlflow server, no authentication, default SQLite backend) exposes the model-registry webhooks API unauthenticated, including a synchronous POST /api/2.0/mlflow/webhooks/{id}/test endpoint that returns the upstream response status and body to the caller. The SSRF guard added in PR #20747 (validatewebhookurl, shipped in 3.10.0) resolves the webhook hostname and rejects non-public IPs, but it is bypassable: delivery follows HTTP redirects (no allowredirects=False) and never pins the validated IP. An attacker hosts a public HTTPS endpoint that passes the guard and returns 302 Location: http://169.254.169.254/... (or http://127.0.0.1:...); MLflow follows it and never re-validates the redirect target. Because /test reflects the response body, this is an unauthenticated full-read SSRF on a default server.

Details Three facts combine:

1. Webhook endpoints are unauthenticated on a default server. The only webhook authorization lives in the optional auth plugin (mlflow/server/auth/init.py, WEBHOOKBEFOREREQUESTHANDLERS), which is not loaded by default.

2. The guard validates but pins nothing — mlflow/utils/validation.py validatewebhookurl:

The resolved IP is never carried into the connection.

CISA’s summary

MLflow contains a server-side request forgery vulnerability that can allow attackers to reach internal or cloud metadata services and receive response_status and response_body.

Weakness type: CWE-918.

Product: MLflow MLflow · Assigned by: GitHub_M · Published: 2026-08-17

Affected and fixed versions

Severity scores

• Attack vector: network (remotely reachable)
• Attack complexity: low
• Privileges required: none (no login)
• User interaction: none
• Scope: changed (can affect other components)

CVSS measures how bad a flaw is if exploited, not how likely exploitation is. Read it next to the exploitation data below.

Exploitation and urgency

• CISA KEV: added 2026-08-19; federal due date 2026-09-02. Known ransomware use: Unknown.
• CISA required action: Apply mitigations in accordance with vendor instructions, ensuring compliance with CISA’s BOD 26-04 Prioritizing Security Updates Based on Risk (see URL in Notes) guidance and CISA’s “Forensics Triage Requirements” (see URL in Notes). Follow applicable BOD 26-04 guidance for cloud services or discontinue use of the product if mitigations are unavailable. Stakeholders are responsible for evaluating each asset's internet exposure and ensuring adherence to BOD 26-04 patching guidelines.
• CISA SSVC assessment: exploitation active, automatable yes, technical impact partial.
• FIRST EPSS (model of 2026-10-21): 9.84% probability of exploitation activity in the next 30 days; percentile 96.

Timeline

• 2026-08-02 — erroneo.us brief: MLflow CVE-2026-64849: Exploited Unauthenticated SSRF
• 2026-08-17 — CVE record published by GitHub_M
• 2026-08-17 — GitHub-reviewed advisory GHSA-7gwp-5pfp-969j published (critical)
• 2026-08-19 — Added to CISA’s Known Exploited Vulnerabilities catalog
• 2026-09-02 — CISA remediation deadline for US federal agencies
• 2026-10-02 — CVE record last updated

Sources and references

Original sources first. Version and fix details should always be confirmed against the vendor or project advisory.

Vendor and project advisories, patches

1. Listed in CISA KEV notes (github.com) github.com
2. Listed in CISA KEV notes (github.com) github.com

Vulnerability records

1. GitHub Advisory GHSA-7gwp-5pfp-969j github.com
2. CVE.org record CVE-2026-64849 cve.org
3. NVD entry CVE-2026-64849 nvd.nist.gov

Government

1. CISA Known Exploited Vulnerabilities catalog entry cisa.gov
2. Listed in CISA KEV notes (cisa.gov) cisa.gov
3. Listed in CISA KEV notes (cisa.gov) cisa.gov

Other references from the CVE record and advisory

1. github.com github.com
2. github.com github.com
3. github.com github.com
4. github.com github.com
5. github.com github.com
6. pypi.org pypi.org

All exploited MLflow vulnerabilities · Critical PyPI advisories · Today’s Security Digest