CVE 6.5 MEDIUM

vLLM Affected by Denial of Service via Unbounded Frame Count in video/jpeg Base64 Processing_CVE-2026-34755

6.5 / 10
MEDIUM
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H

Description

vLLM is an inference and serving engine for large language models (LLMs). From 0.7.0 to before 0.19.0, the VideoMediaIO.load_base64() method at vllm/multimodal/media/video.py splits video/jpeg data URLs by comma to extract individual JPEG frames, but does not enforce a frame count limit. The num_frames parameter (default: 32), which is enforced by the load_bytes() code path, is completely bypassed in the video/jpeg base64 path. An attacker can send a single API request containing thousands of comma-separated base64-encoded JPEG frames, causing the server to decode all frames into memory and crash with OOM. This vulnerability is fixed in 0.19.0.

Basic Information

ID CVE-2026-34755
Source GitHub_M
Published Apr 6, 2026 at 15:38
Modified Apr 6, 2026 at 18:36

Affected Product

Vendor vllm-project
Product vllm
Version >= 0.7.0, < 0.19.0
Affected Versions vllm-project vllm >= 0.7.0, < 0.19.0

CWE Classification

References

💭 Join the Security Discussion

🔒 Your email address will not be published. Required fields are marked *

⚠️ Please be respectful and constructive in your comments. Security discussions should remain professional.