{"id":"CVE-2026-90554","aliases":["GHSA-936p-m5pv-vvjf"],"url":"https://o3.security/vulnerability/CVE-2026-90554","summary":"vLLM before 0.28.0 Denial of Service via audio extraction","details":"vLLM versions >=0.10.2 and <0.28.0 do not apply any audio decode-size or duration limit when extracting audio from video input for NanoNemotronVL models. In nano_nemotron_vl.py, _extract_audio_from_videos calls load_audio_pyav(BytesIO(video_bytes)) without the max_duration_s or max_decode_bytes parameters, so neither VLLM_MAX_AUDIO_DECODE_DURATION_S nor VLLM_MAX_AUDIO_DECODE_BYTES is enforced (unlike the direct audio upload path in AudioMediaIO). When a NanoNemotronVL model is served with use_audio_in_video=True, an attacker who supplies a small, highly compressed video as multimodal input can force the server to allocate gigabytes of memory during audio decoding, resulting in a denial of service. Fixed in vLLM 0.28.0.","published":"2026-09-12T12:08:57.508Z","modified":"2026-09-26T03:48:25.859083213Z","cvss":null,"epss":null,"cisaKev":null,"exploitsKnown":null,"affectedPackages":[],"fix":null,"references":[{"type":"ADVISORY","url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/90xxx/CVE-2026-90554.json"},{"type":"ADVISORY","url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-936p-m5pv-vvjf"},{"type":"ADVISORY","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-90554"},{"type":"ADVISORY","url":"https://www.vulncheck.com/advisories/vllm-before-0.28.0-denial-of-service-via-audio-extraction"}],"provenance":{"sources":["OSV.dev","FIRST.org (EPSS)"],"lastVerified":"2026-09-26T03:48:25.859083213Z"}}