AI and data stack advisories

Severe, 6 weeks2431Projects328

2431 severe, 6 weeks · 328 projects

vLLMGHSA-58v5-2m8f-94pr

vLLM: GLMGA video sampling permits request-driven CPU and memory exhaustion

vLLM

CVE-2026-105760 · Published Oct 5, 2026 · updated Oct 15, 2026

Medium5.3
Fix: upgrade to 0.30.0 or later
GitHub advisory

Summary

The OpenAI-compatible chat endpoint accepts request-level video loader options through media_io_kwargs. A caller can select the GLMGA sampler and provide large fps and max_frames values. GLMGA first constructs an attacker-sized Python index list and then deduplicates it, even when the supplied video contains only a few frames. This work occurs in the shared media-loading executor before frame reads, allowing a tiny valid video and compact JSON options to consume disproportionate CPU time and memory and delay unrelated media requests.

The demonstrated impact is partial denial of service. No memory corruption, data disclosure, or code execution is claimed.

Affected configuration

The server must expose chat completions for a video-capable model and accept request-level media_io_kwargs. The model does not need to use GLMGA by default because the request value overrides the model's loader mapping. The proof of concept uses an inline two-frame video and explicitly selects the OpenCV decoder, so no external media host or GPU decoder is required.

The request is reachable without vLLM credentials when no API key is configured. With API-key authentication enabled, any caller holding a valid key can reach the same path.

Attack surface

A remote caller submits a valid chat-completion request with a small video and the following request-level options:...

Affected versions

PackageAffectedFixed in
vllm
PyPI
>= 0.23.0rc2, < 0.30.00.30.0
Details and references

## Summary The OpenAI-compatible chat endpoint accepts request-level video loader options through `media_io_kwargs`. A caller can select the GLMGA sampler and provide large `fps` and `max_frames` values. GLMGA first constructs an attacker-sized Python index list and then deduplicates it, even when the supplied video contains only a few frames. This work occurs in the shared media-loading executor before frame reads, allowing a tiny valid video and compact JSON options to consume disproportionate CPU time and memory and delay unrelated media requests. The demonstrated impact is partial denial of service. No memory corruption, data disclosure, or code execution is claimed. ## Affected configuration The server must expose chat completions for a video-capable model and accept request-level `media_io_kwargs`. The model does not need to use GLMGA by default because the request value overrides the model's loader mapping. The proof of concept uses an inline two-frame video and explicitly selects the OpenCV decoder, so no external media host or GPU decoder is required. The request is reachable without vLLM credentials when no API key is configured. With API-key authentication enabled, any caller holding a valid key can reach the same path. ## Attack surface A remote caller submits a valid chat-completion request with a small video and the following request-level options: ```json { "media_io_kwargs": { "video": { "video_backend": "glmga", "backend": "opencv", "fps": 500000, "max_frames": 500000 } } } ``` Increasing both numeric values increases temporary list construction and deduplication work even when the final decoded frame set remains unchanged. ## Root cause 1. `ChatCompletionRequest` exposes `media_io_kwargs` as request-controlled nested values: [`protocol.py#L365-L371`](https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/entrypoints/openai/chat_completion/protocol.py#L365-L371). 2. The request options are carried into chat parameters without a numeric work bound on GLMGA's `fps` or `max_frames`: [`protocol.py#L571-L599`](https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/entrypoints/openai/chat_completion/protocol.py#L571-L599). 3. Media options are merged so request values override configured defaults, including `video_backend`: [`connector.py#L577-L605`](https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/multimodal/media/connector.py#L577-L605). 4. The connector selects a registered video loader and runs media loading through the shared executor: [`video.py#L28-L66`](https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/multimodal/media/video.py#L28-L66) and [`connector.py#L44-L47`](https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/multimodal/media/connector.py#L44-L47). 5. GLMGA calculates `extract_t = min(int(duration * fps), max_frames)`, builds a list with that many entries, and deduplicates it before decoding frames: [`video.py#L667-L740`](https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/multimodal/video.py#L667-L740). 6. The OpenCV decoder receives the already-deduplicated indices, so a two-frame file can trigger a large intermediate allocation while producing only two decoded frames: [`opencv.py#L28-L50`](https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/multimodal/video_decoders/opencv.py#L28-L50). The missing invariant is a strict upper bound on sampling work before the index list is constructed. Bounding decoded output after deduplication does not bound the vulnerable intermediate computation. ## Suggested remediation Validate request-level video sampling options before dispatching work to the media executor. Enforce conservative absolute limits for `fps`, `max_frames`, and especially the computed candidate count

CVSS 3.1
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L
Severity from
GitHub (reviewed advisory)
Weakness
CWE-400
Also known as
CVE-2026-105760, PYSEC-2026-4307

More vLLM advisories

All vLLM