# vLLM improper image metadata normalization in image processing

Source: https://junglewise.ai/threats/vllm-improper-image-metadata-normalization-in-image-processing-68504b9d  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-06-17. Updated: 2026-06-18  
License: CC BY 4.0 with a link back

Severity: medium · CVSS 4.8 · Published 2026-06-17

vLLM is an open-source library used to run large language models efficiently. A flaw in how it processes images can cause the AI to misinterpret visual data, such as seeing rotated images or distorted colors where transparency was intended. This could allow an attacker to manipulate the model's output or bypass safety filters by providing specially crafted image files that the AI "sees" differently than a human would.

Technologies: [Red Hat AI Inference Server](https://junglewise.ai/threats/technologies/ai-inference-server). Vendors: [Red Hat](https://junglewise.ai/threats/vendors/red-hat).

## Quick answers

**Is This vulnerability exploited in the wild?** No. Junglewise Threat Intelligence has no report of This vulnerability being exploited in the wild.

**How severe is This vulnerability?** Medium severity, with a CVSS score of 4.8, published 17 June 2026.

**Which products and versions are affected?** vLLM Project vLLM >= 0.11.0, <= 0.23.0; Red Hat Red Hat AI Inference Server 3.

**Is there a fix for This vulnerability?** No fixed version is recorded yet. Check the vendor's advisory in the references below. If you run it, Junglewise drafts a mitigation that fits your setup and applies it once you approve, with senior security engineers behind it.

**How is This vulnerability attacked?** It is attacked over the network, in a software library.

**Am I affected?** Check whether you run Red Hat AI Inference Server. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## Technical details

A vulnerability exists in vLLM's image processing pipeline due to improper handling of EXIF orientation and PNG transparency (tRNS) data. Specifically, the library fails to call ImageOps.exif_transpose after opening images, and it only flattens transparency for RGBA modes, ignoring tRNS chunks in other modes like P, L, or RGB. This results in 'Misinterpretation of Input' (CWE-115), where the model processes distorted or incorrectly oriented pixels. An attacker can exploit this via the network by submitting crafted images to an inference endpoint to cause integrity failures or bypass vision-language model constraints.

## Affected products

- vLLM Project vLLM >= 0.11.0, <= 0.23.0
- Red Hat Red Hat AI Inference Server 3

## Timeline

- 2026-06-17: disclosed: Initial disclosure and CVE assignment
- 2026-06-18: advisory: GitHub advisory published and later withdrawn as a duplicate

## References

- [https://access.redhat.com/security/cve/CVE-2026-12491](https://access.redhat.com/security/cve/CVE-2026-12491)
- [https://bugzilla.redhat.com/show_bug.cgi?id=2489786](https://bugzilla.redhat.com/show_bug.cgi?id=2489786)

## Related threats

- [vLLM improper image metadata handling in image processing](https://junglewise.ai/threats/cve-2026-12491-vllm-improper-image-metadata-handling-in-image-processing)
- [Kludex python-multipart denial of service in multipart header parsing](https://junglewise.ai/threats/cve-2026-42561-kludex-python-multipart-denial-of-service-in-multipart-header)
- [Apache Thrift improper certificate validation in Java TSSLTransportFactory](https://junglewise.ai/threats/cve-2026-41603-apache-thrift-improper-certificate-validation-in-java)
- [Apache Thrift integer overflow in Go TFramedTransport](https://junglewise.ai/threats/cve-2026-41602-apache-thrift-integer-overflow-in-go-tframedtransport)
- [vLLM Denial of Service via unbounded n parameter in OpenAI API server](https://junglewise.ai/threats/cve-2026-34756-vllm-denial-of-service-via-unbounded-n-parameter-in-openai-api)