Vulnerability report

Active exploitation observed Confirmed confidence In CISA KEV

CVE-2026-64849

mlflow Server-Side Request Forgery

Mlflow · >= 3.3.0, < 3.15.0

Watch CVE-2026-64849 Browse KEV feed
Severity
CVSS 9.3 · Critical
Confidence
Confirmed
Exploit status
Observed in sensors
EPSS
9.8%
First observed
23 Sep 2026
Last observed
12 Oct 2026

Decision summary

What security teams need to know first

Direct answers before the deeper technical record.

What it is

CVE-2026-64849 is an unauthenticated vulnerability affecting mlflow. 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...

Is it exploited?

Yes. Previdian sensors observed exploitation attempts with confirmed confidence. Listed in CISA KEV. Also confirmed by third-party sources.

Who is affected?

mlflow >= 3.3.0, < 3.15.0.

What should we do?

Patch immediately, validate internet-facing exposure, and monitor for matching requests.

Overview

MLflow is an open source AI engineering platform for agents, large language models, and machine learning models

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 _validate_webhook_url() 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 response_status and response_body.

This issue is fixed in version 3.15.0.

mlflow Affected
mlflow
>= 3.3.0, < 3.15.0
Published
17 Aug 2026
Exploitation Reported
18 Aug 2026
Attack vector
Remote
Complexity
Low
Privileges
None
User interaction
None

Tags

nuclei_scanner cisa

CVE References

Exploitation evidence

Why Previdian marks this CVE as exploited

Third-party attestation and first-party sensor observation are shown separately so teams can judge the evidence chain.

Exploited in the wild

watchTowr

Recorded 18 Aug 2026

A trusted third party reported exploitation.

Active exploitation observed

Previdian sensor

First observed 23 Sep 2026

Previdian first observed exploitation attempts targeting this vulnerability in our honeypot sensors.

Proof of concept available

GitHub Advisory Database

Recorded 17 Aug 2026

Public scanner or PoC coverage increases practical exploitability.

Known exploited vulnerability sources

Per-source evidence links for KEV attestations are available through the Previdian Pro API.

Learn about Pro API access
Source Added
watchTowr First 2026-08-18 13:31 UTC
CISA 2026-08-19 17:00 UTC
CVE 2026-08-19 17:51 UTC
The Shadowserver 2026-09-18 00:00 UTC
Previdian Sensors 2026-09-23 05:50 UTC

Operational indicators for this CVE are listed under Detection.

Sensor telemetry

First-party evidence of exploitation activity

Previdian sensors recorded exploitation attempts targeting this vulnerability. The cards and chart show volume, unique attackers, and daily activity.

4

Attempts observed

3

Unique attacker IPs

3

Attacker countries

BG · CH · ID

3

Sensors observed

CVE-2026-64849 exploitation attempts over the last 26 days

Daily events observed by Previdian sensors

Updated 17 Oct 2026

0
1
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
2
0
0
0
0
0
22 Sep 28 Sep 4 Oct 10 Oct 17 Oct

First observed 23 Sep 2026 · Last observed 12 Oct 2026

Pro adds sensor region and window summaries. Enterprise adds raw IPs, paths, User-Agents, and payloads.

Start 14-day Pro trial

Detection

Operational artifacts and observed signals

Make the evidence actionable in scanner, SOC, and edge-control workflows.

Observed signals

Request targets
User-Agents
Callback hosts
1
4
0

Request targets and User-Agents available in Pro. Callback host details available in Enterprise.

Virtual patch status

No Previdian virtual patch is currently available. Future rules ship for ModSecurity, Cloudflare, and AWS WAF.

Learn about virtual patches →

Attacker IP Indicators

Attacker IP indicators observed · available in Pro and Enterprise.

Sensor-derived attacker IP indicators are available to Pro and Enterprise accounts under Detection and through the Pro API.

Learn about Pro API access

Scanner Artifacts

Scanner and exploit-framework references linked to this CVE.

Scanner Reference Detected
Nuclei https://github.com/projectdiscovery/nuclei-templates/blob/main/http/cves/2026/CVE-2026-64849.yaml 19 Aug 2026

Risk and context

Severity, weaknesses, and research context

CVSS v3.1

9.3 Critical

Potential damage if exploited. Separate from whether attackers are using it.

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:L/A:N

EPSS

9.8%

Estimated chance of exploitation in the next 30 days. Previdian's warning comes from evidence, not this score.

Weaknesses

Server-Side Request Forgery (SSRF) CWE-918

Recent mention · RecordedFuture

August 2026 CVE Landscape

In August 2026, Insikt Group® identified 73 high-impact vulnerabilities that should be prioritized for remediation, 43 of which had a Very Critical Recorded Future Risk Score. This represents a 14% decrease from last month.

Read full advisory

Potential Proof of Concepts

These PoCs are unverified and could contain malware. Use at your own risk.

isaca0315/CVE-2026-64849-poc-lab

github · Created 2026-09-06 00:25:40 UTC · 0 stars · AI assessment 90%

este laboratorio puede estar bien o mal preguntale a la IA estoy probando pero debe funcionar hahahah

MLflow unauthenticated full-read SSRF in webhook delivery via redirect-follow bypass of _validate_webhook_url guard

pruva · Created 2026-08-23 15:44:23 UTC

Severity: CRITICAL | CVE: CVE-2026-64849 | Package: mlflow/mlflow

zavisco/CVE-2026-64849.yaml

github · Created 2026-08-19 14:16:36 UTC · 0 stars · AI assessment 85%

BiuTrap/CVE-2026-64849

github · Created 2026-08-19 01:21:22 UTC · 0 stars · AI assessment 85%

CVE-2026-64849 PoC

MLflow Webhook SSRF - Unauthenticated Full-Read via Redirect Bypass

nuclei · Created Unknown

https://github.com/mlflow/mlflow/security/advisories/GHSA-7gwp-5pfp-969j

ghsa · Created Unknown · AI assessment 90%

MLflow: Unauthenticated full-read SSRF in webhook delivery: _validate_webhook_url bypassed via unvalidated HTTP redirects (and DNS rebinding)

Timeline

From disclosure to observed exploitation

  1. 05:48 UTC

    Indicators of compromise added (3)

    Indicators of compromise recorded

  2. 05:48 UTC

    Observed by Previdian sensors

    Evidence-backed exploitation signal

  3. 00:00 UTC

    KEV confirmed by The Shadowserver

    Exploitation attested by an external source

  4. 17:51 UTC

    KEV confirmed by CVE

    Exploitation attested by an external source

  5. 17:00 UTC

    Added to CISA KEV

    Listed in the CISA Known Exploited Vulnerabilities catalog

  6. 04:30 UTC

    Nuclei template available

    Scanner coverage available

  7. 13:31 UTC

    Added to Previdian KEV Feed

    High-confidence, third-party attested exploitation

  8. 00:00 UTC

    Added to the Previdian watchlist

    Pro and Enterprise Watch users had 1 day of early warning before CISA KEV. Also 14 hours before Previdian confirmed it as a KEV.

  9. 22:10 UTC

    Public PoC available

    Public proof-of-concept code published

  10. 21:16 UTC

    CVE published

    Vulnerability disclosed publicly

  11. 18:31 UTC

    CVE ID reserved

    Identifier reserved by the CNA

Pro API

Automate this intelligence

Confidence, exploit status, sensor telemetry, PoCs, scanner integrations, mentions, and tags are available programmatically for VM, SOC, and CTI workflows.

Start 14-day Pro trial Read API docs
  • Evidence confidence and provenance
  • First-party sensor telemetry
  • PoC and scanner references
  • Affected versions and enrichment
  • Automation-ready JSON delivery

GET /api/v2/pro/kevs/CVE-2026-64849

Sample API response 
{
  "cve_id": "CVE-2026-64849",
  "confidence": "Confirmed",
  "cvss_score": 9.3,
  "cvss_estimated": false,
  "epss_score": 0.09839,
  "exploit_status": {
    "exploited_in_the_wild": true,
    "active_exploitation_observed": true
  },
  "sensor_telemetry": { "attempts": 4, "sensors": 3 }
}