vLLM leaks a heap address when PIL throws an error

[nixpkgs-security-tracker]

• CVE-2026-22778
• Nixpkgs security tracker issue
• affected package maintainers: cc @CertainLach, @happysalada, @daniel-fahey

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 CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H

• CVSS version: 3.1
• Attack Vector (AV): Network (N)
• Attack Complexity (AC): Low (L)
• Privileges Required (PR): None (N)
• User Interaction (UI): None (N)
• Scope (S): Unchanged (U)
• Confidentiality (C): High (H)
• Integrity (I): High (H)
• Availability (A): High (H)

Affected packages

• vllm (0.8.3@nixos-25.05, 0.11.2@nixos-unstable)
• pkgsRocm.vllm (0.11.2@nixos-unstable)
• python312Packages.vllm (0.8.3@nixos-25.05, 0.11.2@nixos-unstable)
• python313Packages.vllm (0.8.3@nixos-25.05, 0.11.2@nixos-unstable)
• pkgsRocm.python3Packages.vllm (0.11.2@nixos-unstable)

Additional comment

Upstream fix: https://github.com/vllm-project/vllm/releases/tag/v0.14.1
Upstream advisory: https://github.com/vllm-project/vllm/security/advisories/GHSA-4r2x-xpjr-7cvv

Unstable fix: https://github.com/NixOS/nixpkgs/pull/483505