CVE-2026-22778CRITICAL 9.8EPSS p60.9%
CVE-2026-22778CVE-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 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.
Scoring
| CVSS 3.1 | 9.8 (CRITICAL) |
| Vector | CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H |
| EPSS | 1.08% probability of exploitation · percentile 60.9% · 2026-06-19T12:03:05Z |
| Published | 2026-02-02 |
| Last modified | 2026-02-23 |
Underlying weaknesses· 1
References
1
| Type | Target | Confidence | Tier |
|---|---|---|---|
| Weakness | Insertion of Sensitive Information into Log Filecwe-532 | 0% | live |
Related by meaning· 6
Nearest entities by semantic similarity across the cs-graph corpus.