Bug 2535557 (CVE-2026-57173) - CVE-2026-57173 vllm: vLLM: Denial of Service via unauthenticated audio decompression
Summary: CVE-2026-57173 vllm: vLLM: Denial of Service via unauthenticated audio decomp...
Keywords:
Status: NEW
Alias: CVE-2026-57173
Product: Security Response
Classification: Other
Component: vulnerability
Version: unspecified
Hardware: All
OS: Linux
medium
medium
Target Milestone: ---
Assignee: Product Security DevOps Team
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2026-09-16 16:53 UTC by OSIDB Bzimport
Modified: 2026-09-24 12:15 UTC (History)
4 users (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed:
Embargoed:


Attachments (Terms of Use)

Description OSIDB Bzimport 2026-09-16 16:53:52 UTC
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.


Note You need to log in before you can comment on or make changes to this bug.