CVE-2026-12491Public exploitNetwork attack vector

Model input distortion via unnormalized image metadata

Published Jun 17, 2026 · Updated Jun 17, 2026

Input misinterpretation in vLLM 0.11.0 through 0.23.0 allows remote attackers to alter model-visible content via crafted images. The image.py conversion path omits EXIF transposition and converts non-RGBA PNG tRNS modes directly to RGB without first compositing transparency. Exploitation requires image input to a multimodal endpoint; displayed pixels can differ from model input, altering reasoning or disrupting downstream processing.

CVSS severity4.8
Medium
EPSS probability0.24%
Next 30 days · Sep 16, 2026
Known exploitationUnconfirmed
Based on sourced intelligence
Hinoki checkNot available
Coverage for this vulnerability

See if you're affected

Explore vulnerability checks for your environment with Hinoki.

Book a demo

Summary

What happened

Input misinterpretation in vLLM 0.11.0 through 0.23.0 allows remote attackers to alter model-visible content via crafted images. The image.py conversion path omits EXIF transposition and converts non-RGBA PNG tRNS modes directly to RGB without first compositing transparency. Exploitation requires image input to a multimodal endpoint; displayed pixels can differ from model input, altering reasoning or disrupting downstream processing.

The record

CVE
CVE-2026-12491
Published
Jun 17, 2026
Updated
Jun 17, 2026
Vendor
389 Directory Server
Product
Red Hat OpenShift AI
Classifications
CWE-436, CWE-115
Attack vector
network
Privileges
unauthenticated

Timeline

How it unfolded

  1. Jun 17, 2026CVE publishedPublication date reported by the CVE source.
  2. Jun 17, 2026Record updatedLatest update available in the CVE record.

Exploitability

Present is not the same as exploitable

Compare your product and version with the public record. A matching version still requires validation against your environment.

Is a vulnerable build present?

Affected versions are unavailable in this record. Check the vendor advisory for version and patch details.

What conditions does exploitation require?

Attack vectornetwork
Required privilegesunauthenticated

What is affected?

389 Directory Server · Red Hat OpenShift AIAffected versions are unavailable in this record.
389 Directory Server · Red Hat Enterprise Linux AIAffected versions are unavailable in this record.
389 Directory Server · Red Hat AI Inference ServerAffected versions are unavailable in this record.

Published CVSS scores

4.8Red HatCVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:L

CVSS describes severity. EPSS estimates exploitation probability.

Attacks

What attackers are doing with it

Daily unique IPs observed by Shadowserver honeypots for known exploited vulnerabilities (KEVs). Missing observations do not establish an absence of attacks.

Daily unique IPsNo honeypot observations are available for this CVE in the selected window.

No observations available

Sep 9, 2026Sep 15, 2026
Latest reporting daySep 15, 2026
Latest daily unique IPsUnavailable
Prior 30-day averageUnavailable
SourceShadowserver honeypots (KEV)
Vectornetwork
Privilegesunauthenticated
Known exploitationUnconfirmed
Public exploitPublished

Weakness, pattern, technique

CWE-436Interpretation Conflict
CWE-115Misinterpretation of Input

Public exploit references

Labels summarize the accepted research assessment. They do not indicate a test against your environment.

Technologies

Your stack

See the directory against your own environment.

Your stack

Check the software in your environment

Book a demo to see how Hinoki identifies affected software and validates exploitability in your environment.

Book a demo