CVE-2026-114053

critical

Description

vLLM from 0.7.3 before 0.31.0 forwards the request-supplied mm_processor_kwargs.code_revision to Transformers AutoProcessor.from_pretrained, letting clients select which revision of a model's dynamic processor code is imported when trust_remote_code is enabled. An attacker controlling a revision in the served model repository can send a multimodal request to the unauthenticated /tokenize endpoint to execute arbitrary Python in the vLLM API process.

References

https://www.vulncheck.com/advisories/vllm-0.7.3-before-0.31.0-rce-via-mm-processor-kwargs-code-revision

https://github.com/vllm-project/vllm/security/advisories/GHSA-h3rc-6mm3-gc2m

https://github.com/vllm-project/vllm/commit/570735520942b6428ae98f1eb442c6d88adae55e

https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-107004

Details

Source: Mitre, NVD

Published: 2026-10-21

Updated: 2026-10-27

Named Vulnerability: GHSA-h3rc-6mm3-gc2mNamed Vulnerability: GHSA-f623-hjm8-5wg5

Risk Information

CVSS v2

Base Score: 7.6

Vector: CVSS2#AV:N/AC:H/Au:N/C:C/I:C/A:C

Severity: High

CVSS v3

Base Score: 8.1

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

Severity: High

CVSS v4

Base Score: 9.2

Vector: CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N

Severity: Critical

EPSS

EPSS: 0.00462