CVE-2026-115915: vLLM insecure image parsing command execution vulnerability

# CVE-2026-115915: vLLM insecure image parsing command execution vulnerability

https://gcdradar.vercel.app/incidents/cve-2026-115915-vllm-insecure-image-parsing-command-executio · AIR-2026-1540 · closed 2026-10-26

- Lane: Patch now
- Severity: 2 (Disclosed vulnerability): Disclosed vulnerability with a patch available and no confirmed exploitation.
- Fix status: Patched
- First reported: 2026-10-26
- AI role: Attack on AI
- Products: vLLM
- CVE ids: CVE-2026-115915
- CVSS: 8.1

## Summary

Improper restriction of decoder selection during image parsing allows attackers to dispatch EPS payloads. This can lead to arbitrary command execution if vulnerable underlying components like Ghostscript or Pillow are present.

## What to do

Update to vLLM 0.32.0 or later.

## Affected and fixed versions

- Affected: vLLM < 0.32.0
- Fixed in: vLLM 0.32.0

## Latest updates

- 2026-10-28: GitHub advisory published (https://github.com/advisories/GHSA-h3rc-j82p-2c3h)
- 2026-10-26: CVE-2026-115915: vLLM is an inference and serving engine for large language models. (https://nvd.nist.gov/vuln/detail/CVE-2026-115915)

## Sources

- 2026-10-26 NVD: CVE-2026-115915: vLLM is an inference and serving engine for large language models. https://nvd.nist.gov/vuln/detail/CVE-2026-115915
- 2026-10-28 GitHub Advisory Database: CVE-2026-115915: vLLM: Untrusted EPS inputs can trigger denial of service or Ghostscript RCE https://github.com/advisories/GHSA-h3rc-j82p-2c3h

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.