CVE-2026-57173: vLLM: Unauthenticated audio decompression-bomb DoS in /v1/chat/completions

Vulnerability summary

CVE-2026-57173 is a medium (CVSS 6.5) vulnerability in vllm-project vllm. EPSS estimates a 0.73% chance of exploitation in the next 30 days, higher than 53.7% of scored vulnerabilities.

Timeline

1. 16-09-2026
   PublishedVulnerability first disclosed.
2. 18-09-2026
   Last modifiedVulnerability information updated.

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.

• CWE-770Allocation of Resources Without Limits or ThrottlingThe product allocates a reusable resource or group of resources on behalf of an actor without imposing any intended restrictions on the size or number of resources that can be allocated.