Описание
vLLM: Unauthenticated audio decompression-bomb DoS in /v1/chat/completions
Summary
The audio decode-duration guard (max_duration_s, env VLLM_MAX_AUDIO_DECODE_DURATION_S, default 600s) that protects against audio decompression-bomb DoS is wired into only the speech-to-text path (/v1/audio/transcriptions). The chat audio path (/v1/chat/completions, input_audio content parts) calls the same decoder with no limit, so an unauthenticated client can submit a few-KB compressed audio file that expands to multiple GB of float32 PCM at decode time, OOM-killing the worker. This is a distinct sibling of CVE-2026-5497 (video frame-count bomb, VideoMediaIO.load_base64) and GHSA-pq5c-rjhq-qp7p (image) in the same media subsystem.
Verified against main at HEAD d78650c (2026-06-16); applicable to the latest release v0.23.0.
Details
The guard rejects long audio during decode (before allocation), implemented in vllm/multimodal/media/audio.py:
load_audio_pyav— metadata reject (~82-98) and live sample-count reject (~129-136)load_audio_soundfile— frames reject (~165-174)
All are gated on if max_duration_s is not None.
It is passed in exactly one place — the transcription serving layer:
The chat path never threads it:
Unauthenticated reachability chain (chat):
parse_input_audio (chat_utils.py) -> parse_audio -> connector.fetch_audio -> AudioMediaIO._load_data_url -> load_base64 -> load_bytes -> load_audio(..., sr=None). The connector never passes max_duration_s, and inline data: URLs need no HTTP fetch (so VLLM_AUDIO_FETCH_TIMEOUT does not bound them). The OpenAI-compatible server has no auth by default (auth only when --api-key / VLLM_API_KEY is set).
Impact
Unauthenticated remote denial of service (availability) via memory amplification on a default-no-auth endpoint, on any deployment serving an audio-capable model. Same class and impact as the sibling CVE-2026-5497 (video). CWE-770 / CWE-409.
Fix
A fix was introduced in this MR: https://github.com/vllm-project/vllm/pull/45908
Ссылки
- https://github.com/vllm-project/vllm/security/advisories/GHSA-hcwq-8wjf-3gcr
- https://nvd.nist.gov/vuln/detail/CVE-2026-57173
- https://github.com/vllm-project/vllm/pull/45908
- https://github.com/vllm-project/vllm/commit/3d20275bb4d434f53055c3c0b645fd8bb072965e
- https://github.com/vllm-project/vllm/releases/tag/v0.24.0
Пакеты
vllm
<= 0.23.0
0.24.0
Связанные уязвимости
A flaw was found in vLLM, an inference and serving engine for large language models. An unauthenticated client can exploit this by submitting a small, compressed audio input that expands into a very large memory allocation. This bypasses existing duration guards, leading to an out-of-memory condition and causing a worker crash. This vulnerability results in a Denial of Service (DoS) for deployments serving audio-capable models.
vLLM is an inference and serving engine for large language models. Prior to 0.24.0, the input_audio handling path for /v1/chat/completions calls AudioMediaIO.load_bytes or AudioMediaIO.load_file without passing VLLM_MAX_AUDIO_DECODE_DURATION_S to the shared audio decoder. An unauthenticated client can therefore submit a small compressed audio input that expands into a very large float32 PCM allocation, bypassing the duration guard already used by /v1/audio/transcriptions and causing an out-of-memory worker crash. Inline data URLs reach this path without being bounded by VLLM_AUDIO_FETCH_TIMEOUT. The issue affects deployments serving an audio-capable model, and authentication changes only the deployment-specific reachability. This issue is fixed in version 0.24.0.
vLLM is an inference and serving engine for large language models. Pri ...