VulnerabilityModified

CVE-2026-22778

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.

CRITICAL 9.8EPSS 18.5%

Does this matter?

EPSS puts the probability of exploitation in the next 30 days at 18.5%, higher than 97% of all known CVEs. Patch or mitigate before the next change window.

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.

CVSS 3.1: 9.8 CRITICALCVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
EPSS: 18.54% probability · 97th percentile
Public exploits: None in Exploit-DB
CISA KEV: Not listed
Weakness: CWE-532, CWE-209
Affected: vllm/vllm
Source: security-advisories@github.com

EPSS trend

EPSS moved from 10.5% on 27 September 2026 to 18.5% on 26 October 2026. The dashed line is the ten per cent mark our analysts treat as “patch before the next change window”.

References

Source: NVD record, EPSS from FIRST.org, KEV from CISA, exploits from Exploit-DB. Refreshed daily. Download this record as JSON.