CVE-2026-64849
CVE-2026-64849 — MLflow: Unauthenticated full-read SSRF in webhook delivery: _validate_webhook_url bypassed via unvalidated HTTP redirects (and DNS rebinding)
MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. 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.
Published Updated Sources: NVD, mlflow (CNA), CISA, FIRST EPSS, GitHub, Red Hat (VEX), SOCRadar CTI
Triage
Is it exploited, how likely is exploitation, what does it touch, and how severe do the scoring sources call it.
Exploitation
Confirmed
CISA KEV, listed Aug 19, 2026
EPSS
16%
ahead of 97% of scored CVEs
CVSS base
9.3
critical · 2 sources
CISA SSVC assessment
Three decision points CISA publishes for the CVEs it assesses · SSVC 2.0.3. A stakeholder decision, not a severity score.
Exploitation
Active
none · proof-of-concept · active
Automatable
Yes
can an attacker script all four kill-chain steps
Technical impact
Partial
partial · total control of the vulnerable component
Remediation
The vendor's own words where we have them.
Upgrade mlflow to a fixed release. Apply vendor patches per advisory and restrict external exposure of the affected component until patched.
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.
Federal deadline Sep 2, 2026 — in 3 days
Red Hat VEX · CVE-2026-64849
vendor_fix: For Red Hat OpenShift AI 3.4.4 see the following documentation, which will be updated shortly for this release, for important instructions on how to upgrade your cluster and fully…
workaround: To reduce the attack surface for this vulnerability, restrict network access to the MLflow server. Implement firewall rules or network access controls to limit connectivity to the…
First 24 hours
Ordered from the record's own fields — exposure first, because you cannot patch what you have not found.
- Identify exposed assets running affected vendor/product/version combinations.
- Treat as emergency remediation because CISA KEV status is present.
- Search available logs for exploit probes, errors, authentication anomalies, or suspicious child processes matching the vulnerability class.
Affected scope
Vendor, product and version as the advisories word them.
| Vendor | Product | Versions | Status |
|---|---|---|---|
| mlflow | mlflow | < 3.15.0 | Vulnerable |
Attack characteristics
The CVSS vector, decoded. It describes the attack, not your exposure to it.
Availability
None
Confidentiality
High
Integrity
Low
Scope
Changed
Attack Complexity
Low
Attack Vector
Network
Privileges Req
None
User Interaction
None
Every base score collected
Sources score independently and disagree; each row says who scored it and under which version.
| Score | Version | Severity | Expl. | Impact | Source |
|---|---|---|---|---|---|
| 9.3 | CVSS 3.1 | critical | 3.9 | 4.7 | GitHub (CNA), mlflow (CNA) |
| 8.5 | CVSS 3.1 | high | — | — | Red Hat (VEX) |
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:L/A:N
What this weakness leads to
MITRE's own consequences and mitigations for the weakness class — the authority's wording, not guidance derived from the CVSS vector.
CWE-918 · Server-Side Request Forgery (SSRF)
- Read Application Data
- Execute Unauthorized Code or Commands::
Weakness & attack patterns
- CWE-918Server-Side Request Forgery (SSRF)
Inferred from the weakness class — not observed against this CVE.
CISA KEV details
CISA lists a CVE here once it has evidence of exploitation against real targets.
Listed
Aug 19, 2026
Federal deadline
Sep 2, 2026
in 3 days
Ransomware use
Unknown
CISA records ransomware use only for campaigns it has confirmed, so “Unknown” means unrecorded, not ruled out. The deadline binds US federal agencies; for everyone else it is a date to argue against.
Public exploit
Capability, not use: code existing is a different claim from anyone running it.
Repositories
3
Exploit entries
2
Exploit-DB and CTI exploit indexes
Social mentions
Posts naming this CVE. Attention, not evidence of exploitation — it spikes on disclosure and decays whether or not anything is exploited.
Critical MLflow SSRF vulnerability being actively exploited. CVE-2026-64849 carries a CVSS score of 9.3 and can allow attackers to reach internal cloud metadata services, potentially exposing cloud credentials and secrets. CISA has added it to the KEV catalog.
@webenstein_· Aug 27, 2026
CVE-2026-64849 is more than “MLflow has an SSRF.” A webhook feature can inherit the tracking server’s network trust and reach internal/cloud metadata endpoints. Patch to 3.15.0+ and ask what the server is allowed to reach.
@InfosecDotWatch· Aug 26, 2026
CVSS 9.3: MLflow CVE-2026-64849 SSRF steals cloud IAM creds. No login needed. Active exploitation confirmed. CISA KEV. Federal deadline Sep 2. Patch to 3.15.0. Enable IMDSv2. Rotate creds. Full breakdown: https://t.co/DkdbC1DFrN #MLflow #CISAkev
@DecryptionDigst· Aug 25, 2026
DEEP DIVE: CVE-2026-64849 is an unauthenticated SSRF in MLflow, CISA KEV since Aug 19. The webhook test endpoint reflects the internal response to the caller, so it reads your network rather than just probing it. The fix shipped in 3.15.0 six weeks before the advisory. https://t.co/vyg1zoJ5Z0
@DailyCVEBrief· Aug 24, 2026
🚨 CVE-2026-64849: Affecting versions prior to 3.15.0. An attacker can bypass webhook URL validation and read internal or cloud metadata responses. A working Goby EXP has been released. Link: https://t.co/6h35NysxDN https://t.co/0E6RW0ZKmj
@GobySec· Aug 24, 2026
28 more posts on this CVEShowing 5 of 33 · Who is discussing this CVE, and how loudlyUnlock freeDetection
Read off the CVSS vector and the weakness class. Starting points, not rules we have tested.
- Search application, proxy, and WAF logs for requests touching /api/2.0/mlflow/webhooks/, /test, /utils/validation.py, /webhooks/delivery.py.
- Prioritize edge telemetry for network-reachable mlflow.
- Monitor for scanner or exploit-pattern traffic after 5 public PoC repositories were reported.
Timeline
What happened to this CVE, newest first — with the readings a source repeats on a schedule counted underneath rather than listed.
- 2026
Added · CVSS 3.1 9.3 (AV:N)
Aug 20, 2026 · NVD
Added to CISA KEV catalog
Aug 19, 2026 · CISA
Cve is exploited BiuTrap/CVE-2026-64849
Aug 19, 2026 · SOCRadar CTI
Cve is exploited codeb0ssx/CVE-2026-64849-PoC
Aug 19, 2026 · SOCRadar CTI
Initial · CVE published
Aug 17, 2026 · NVD
Source activity
Readings a source repeats on a schedule, counted rather than listed.
- 11×TweetsAug 19, 2026 – Aug 28, 2026 · SOCRadar CTI
- 7×SvrsAug 19, 2026 – Aug 28, 2026 · SOCRadar CTI
- 6×ModifiedAug 17, 2026 – Aug 20, 2026 · SOCRadar CTI
- 3×Github ReposAug 18, 2026 – Aug 19, 2026 · SOCRadar CTI
- 1×CVE appeared in the Cve Trends.Aug 19, 2026 · SOCRadar CTI
References
13 on the record
- github.com/mlflow/mlflow/commit/ba949522477cbd5915aa55d29b0cfad7d5ddf939
Patch, Mitre, NIST
- github.com/mlflow/mlflow/issues/24179
Exploit, Issue Tracking, Mitigation, Mitre, NIST, GITHUB, Vendor Advisory
- github.com/mlflow/mlflow/pull/24258
Issue Tracking, Patch, Mitre, NIST, Vendor Advisory
- github.com/mlflow/mlflow/releases/tag/v3.15.0
Release Notes, Mitre, NIST
- github.com/mlflow/mlflow/security/advisories/GHSA-7gwp-5pfp-969j
Exploit, Vendor Advisory, Mitre, NIST, GITHUB
- www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-64849
US Government Resource, NIST
- access.redhat.com/errata/RHSA-2026:60520
Vendor Advisory
- access.redhat.com/security/cve/CVE-2026-64849
Vendor Advisory
Coverage
10 stories from the feeds we poll
Hackers Are Already Exploiting a Critical Flaw in AI Tool MLflow - kobaran.com
kobaran.com · Aug 24, 2026
Falha do MLflow já está sendo atacada
cisoadvisor.com.br · Aug 20, 2026
CISA alerta para exploração ativa de falha crítica no MLflow
cisoadvisor.com.br · Aug 20, 2026
U.S. CISA adds an MLflow flaw to its Known Exploited Vulnerabilities catalog - Security Affairs
securityaffairs.com · Aug 20, 2026
MLflow Vulnerability Exploited for Cloud Credential Theft - SecurityWeek
securityweek.com · Aug 20, 2026
5 more reports on this CVEShowing 5 of 10 · Vendor research and security press coverageUnlock freeElsewhere on this site
- mlflowevery CVE for this vendor
- August 2026what else CISA listed
- SSRFsame class
- CWE-918other pages naming this weakness
Not in any source we poll
Listed rather than left blank: an empty field and an unmeasured one look identical on screen, and only one is a reason to look elsewhere.
- No confirmed IOCs, IP addresses, domains, file hashes, or malware artifacts supplied.
- No organization-specific asset inventory, compensating-control status, or patch deployment evidence supplied.
- No exploit packet captures, log samples, or incident case IDs supplied.
Answered from this record1
What should defenders know first?
CVE-2026-64849 is MLflow: Unauthenticated full-read SSRF in webhook delivery: _validate_webhook_url bypassed via unvalidated HTTP redirects (and DNS rebinding), a critical vulnerability affecting mlflow from mlflow. The current evidence lists it in CISA KEV, and the exploit status is: Yes: CISA KEV signal present. Public exploit evidence is: 5 public PoC repositories reported; 0 marked weaponized in current dataset. The affected-version evidence is listed in the key facts and affected products tables. Defenders should first verify whether exposed or business-critical assets run those versions, then apply vendor patches or mitigations, restrict reachable attack surface, and preserve logs for detection review. CVSS 9.3 describes technical severity, while EPSS 16% helps estimate near-term exploit likelihood; neither replaces asset context. Unknown fields should remain explicit in tickets, and threat actor, IOC, victimology, or payload claims should not be added unless a cited source supports them. Monitor CISA KEV, vendor advisories, NVD changes, public PoC repositories, and internal telemetry for update triggers.