{"id":"CVE-2026-90554","aliases":["EUVD-2026-76872","GHSA-936p-m5pv-vvjf","GHSA-m39m-2pqx-3vm2"],"status":"published","title":"vLLM before 0.28.0 Denial of Service via audio extraction","description":"vLLM versions \u003e=0.10.2 and \u003c0.28.0 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.","severity":{"version":"4.0","score":6.9,"vector":"CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N","level":"medium","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-11","score":0.00201,"percentile":0.10053,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":false},"cwes":["CWE-400"],"published":"2026-09-12T12:08:57Z","modified":"2026-10-28T02:31:09Z","products":[{"vendor":"vllm","product":"vllm"},{"vendor":"vllm-project","product":"vllm"}],"references":[{"url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-936p-m5pv-vvjf","tags":["vendor-advisory","Vendor Advisory","advisory"]},{"url":"https://www.vulncheck.com/advisories/vllm-before-0.28.0-denial-of-service-via-audio-extraction","tags":["third-party-advisory","Third Party Advisory","advisory"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-90554","tags":["advisory"]},{"url":"https://github.com/advisories/GHSA-m39m-2pqx-3vm2"},{"url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/90xxx/CVE-2026-90554.json","tags":["advisory"]}],"sources":{"cve":[{"id":"CVE-2026-90554","title":"vLLM before 0.28.0 Denial of Service via audio extraction","description":"vLLM versions \u003e=0.10.2 and \u003c0.28.0 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.","status":"published","published":"2026-09-12T12:08:57.508Z","modified":"2026-09-14T18:16:18.655Z","cvss":[{"version":"4.0","score":6.9,"vector":"CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N","level":"medium","assessor":"cna"},{"version":"3.1","score":6.2,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"medium","assessor":"cna"}],"cwes":["CWE-400"],"updated_at":"2026-11-02T13:45:47.885058Z"}],"euvd":[{"id":"EUVD-2026-76872","description":"vLLM versions \u003e=0.10.2 and \u003c0.28.0 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.","status":"published","published":"2026-09-12T12:08:57Z","modified":"2026-09-14T18:16:18Z","cvss":[{"version":"4.0","score":6.9,"vector":"CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N","level":"medium"}],"updated_at":"2026-11-02T17:28:19.730872Z"}],"ghsa":[{"id":"GHSA-m39m-2pqx-3vm2","title":"vLLM versions \u003e=0.10.2 and \u003c0.28.0 do not apply any audio decode-size or duration limit when...","description":"vLLM versions \u003e=0.10.2 and \u003c0.28.0 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.","status":"published","published":"2026-09-12T15:31:50Z","modified":"2026-09-12T15:31:56Z","cvss":[{"version":"4.0","score":6.9,"vector":"CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X","level":"medium"},{"version":"3.1","score":6.2,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"medium"}],"cwes":["CWE-400"],"updated_at":"2026-11-02T15:11:45.662784Z"}],"nvd":[{"id":"CVE-2026-90554","description":"vLLM versions \u003e=0.10.2 and \u003c0.28.0 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.","status":"published","published":"2026-09-12T13:16:54.04Z","modified":"2026-09-24T20:28:13.59Z","cvss":[{"version":"4.0","score":6.9,"vector":"CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X","level":"medium","assessor":"secondary"},{"version":"3.1","score":6.2,"vector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"medium","assessor":"secondary"},{"version":"3.1","score":7.5,"vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H","level":"high","assessor":"primary"}],"cwes":["CWE-400"],"updated_at":"2026-11-02T14:12:46.62721Z"}],"osv":[{"id":"CVE-2026-90554","title":"vLLM before 0.28.0 Denial of Service via audio extraction","description":"vLLM versions \u003e=0.10.2 and \u003c0.28.0 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.","status":"published","published":"2026-09-12T12:08:57.508Z","modified":"2026-10-28T02:31:09.24454092Z","cvss":[{"version":"4.0","score":6.9,"vector":"CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N","level":"medium"}],"cwes":["CWE-400"],"updated_at":"2026-11-02T15:45:20.269367Z"}]}}