CVE-2026-130686

high

Description

vLLM from 0.6.0 before 0.33.0 lacks a request-wide media budget, allowing remote inference clients to exhaust frontend resources by submitting many remote image URLs in one request. Attackers can send a compact request to /v1/messages or /v1/responses that forces downloading and decoding of every image before prompt-length admission, exhausting memory, CPU, and connections.

References

https://github.com/vllm-project/vllm/security/advisories/GHSA-5r3c-5x4v-8gjh

https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-128760

Details

Source: Mitre, NVD

Published: 2026-11-18

Updated: 2026-11-18

CVSS v3

Base Score: 7.5

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

Severity: High

CVSS v4

Base Score: 8.7

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

Severity: High

EPSS

EPSS: 0.00411