Analysis and contextual insights are available on OpenCVE Cloud.
No vendor fix or workaround currently provided.
Additional remediation guidance may be available on OpenCVE Cloud.
Tracking
Sign in to view the affected projects.
No advisories yet.
Wed, 16 Sep 2026 16:45:00 +0000
| Type | Values Removed | Values Added |
|---|---|---|
| Description | 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. | |
| Title | vLLM: Unauthenticated audio decompression-bomb DoS in /v1/chat/completions | |
| Weaknesses | CWE-770 | |
| References |
| |
| Metrics |
cvssV3_1
|
Subscriptions
No data.
Status: PUBLISHED
Assigner: GitHub_M
Published:
Updated: 2026-09-16T16:34:39.144Z
Reserved: 2026-06-24T01:47:55.285Z
Link: CVE-2026-57173
No data.
No data.
No data.
OpenCVE Enrichment
No data.
-
CWE-770
Allocation of Resources Without Limits or Throttling