CVE-2026-22778

Description

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 vLLMs 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.

Risk Information

Base Score
9.8
MODERATE
Vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
EPSS Score
Exploitation Probability
0.084

Associated Vulnerability

VulnerabilityOS Platform
Vulnerabilities CVE-2026-22778,CVE-2026-24779 are fixed in Python-vllm 0.14.1Windows
Vulnerabilities CVE-2026-22778,CVE-2026-24779 are fixed in Python-vllm for linux 0.14.1Linux

Patch Details

No records found

References

https://nvd.nist.gov/vuln/detail/CVE-2023-1234
https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2023-1234