SECURELYN
Free ScanFree consultation

CVE-2026-22778

vLLM is an inference and serving engine for large language models (LLMs). From 0.8.3 to before 0.14.1, when an invalid image is sent to vLLM's multimodal endpoint, PIL throws an error. vLLM returns this error to the client, leaking a heap address. With this leak, we reduce ASLR from 4 billion guesses to ~8 guesses. This vulnerability can be chained a heap overflow with JPEG2000 decoder in OpenCV/FFmpeg to achieve remote code execution. This vulnerability is fixed in 0.14.1.

Scores

9.8
CVSS 9.8 out of 10, critical severity.
16.7%
EPSS 16.7% chance of exploitation in the next 30 days.

The facts

Published
2 February 2026
Last changed
15 July 2026
NVD status
Modified
Severity
Critical, CVSS 9.8 out of 10 (version 3.1)
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
Weakness
Affects
  • vllm vllm
Known exploited
Not in the CISA Known Exploited Vulnerabilities catalogue in this build.
EPSS
16.7% chance of exploitation activity in the next 30 days (percentile 97).

Is there a Securelyn advisory?

Not yet. Securelyn writes plain-English advisories for the flaws that matter most to businesses. Subscribe to the digest to hear about them.

References