Advisory record · GHSA-6w67-hwm5-92mq

LMDeploy has Server-Side Request Forgery (SSRF) via Vision-Language Image Loading

CVSS 7.5 HighCriticalEPSS 0.45 · 98th percentile

Critical, because:

  • CVSS score 7.5 (7.0 or more) and EPSS 0.45 (0.1 or more): a high score and a real chance of exploitation in the next 30 days.

How we decide what’s critical

Published by the GitHub Advisory Database at 21 Apr 2026, 15:04 UTC. agentprime has not analysed this yet, so whether it affects any app is not known.

PackageEcosystemAffectedFixed in
lmdeployPyPI<=0.12.2No fixed version yet
Identifiers
GHSA-6w67-hwm5-92mq · CVE-2026-33626 · PYSEC-2026-2607
Weakness (CWE)
CWE-918
Published
21 Apr 2026
Updated
13 Jul 2026

Machine-written summary

Summary written by Claude Haiku from the source advisory. Not reviewed by a person.

LMDeploy versions up to and including 0.12.2 are affected by a server-side request forgery in the vision-language image loading code, where URLs are fetched without checking for internal or private addresses. An attacker who can send requests to a deployed server with a vision-language model can make it fetch internal resources, such as cloud metadata services, and so reach data not exposed to the internet. The source lists no fixed version, though it states that versions prior to 0.12.3 are affected.

Written 6 Nov 2026.

Not analysed yet

We haven’t published our analysis of this Advisory yet. Once a person has reviewed it, this page will say what’s vulnerable, what has to be true for your app to be affected, and how to check.

Severity and exploitation

ScoreRatingCVSSVectorSource
7.5 High 3.1 CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N CNA (GitHub_M)
  • CISA KEV Not listed.
  • EPSS 0.45, the 98th percentile, on 14 Nov 2026: FIRST’s estimate of the chance of exploitation in the next 30 days.

Timeline

  1. 31 Oct 2026, 14:59 UTC First seen by agentprime
  2. 31 Oct 2026, 14:59 UTC Confirmed by OSV

Sources

Facts on this page come from the GitHub Advisory Database (CC-BY 4.0), OSV and the other sources listed, as published. We don’t reproduce full advisory descriptions, which can contain exploit code.