{"id":"CVE-2026-34760","aliases":["GHSA-6c4r-fmh3-7rh8","PYSEC-2026-2299"],"summary":"vLLM: Downmix Implementation Differences as Attack Vectors Against Audio AI Models","published_at":"2026-04-02T18:59:49.638Z","modified_at":"2026-07-17T17:00:52.155049Z","products":[],"rating":{"severity":"high","score":7.1,"vector":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:L","attack_path":"internet_authenticated","user_interaction":false,"malicious":false,"red_hat":"Moderate","vendor_ratings":[{"source":"Red Hat","rating":"Moderate"}]},"fix_available":false,"kev":{"listed":false,"ransomware":false},"public_exploit":false,"exploits":[],"epss":{"score":0.0028299998957663774,"percentile":0.21230000257492065,"date":"2026-11-19T00:00:00Z"},"priority":8.1,"details":"vLLM is an inference and serving engine for large language models (LLMs). From version 0.5.5 to before version 0.18.0, Librosa defaults to using numpy.mean for mono downmixing (to_mono), while the international standard ITU-R BS.775-4 specifies a weighted downmixing algorithm. This discrepancy results in inconsistency between audio heard by humans (e.g., through headphones/regular speakers) and audio processed by AI models (Which infra via Librosa, such as vllm, transformer). This issue has been patched in version 0.18.0.","affected_stack":[],"affected_other":[{"ecosystem":"CVE","package":"vllm-project/vllm","versions":[">= 0.5.5, < 0.18.0"],"provider":"cvelistv5","source":"CVE-2026-34760","relation":"self"},{"ecosystem":"NVD","package":"vllm:vllm","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0.5.5"},{"fixed":"0.18.0"}]}],"fixed":["0.18.0"],"provider":"nvd","source":"CVE-2026-34760","relation":"self"},{"ecosystem":"PyPI","package":"vllm","purl":"pkg:pypi/vllm","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0.5.5"},{"fixed":"0.18.0"}]}],"versions":["0.10.0","0.10.1","0.10.1.1","0.10.2","0.11.0","0.11.1","0.11.2","0.12.0","0.13.0","0.14.0","0.14.1","0.15.0","0.15.1","0.16.0","0.17.0","0.17.1","0.5.5","0.6.0","0.6.1","0.6.1.post1","0.6.1.post2","0.6.2","0.6.3","0.6.3.post1","0.6.4","0.6.4.post1","0.6.5","0.6.6","0.6.6.post1","0.7.0","0.7.1","0.7.2","0.7.3","0.8.0","0.8.1","0.8.2","0.8.3","0.8.4","0.8.5","0.8.5.post1","0.9.0","0.9.0.1","0.9.1","0.9.2"],"fixed":["0.18.0"],"provider":"osv","source":"GHSA-6c4r-fmh3-7rh8","relation":"self"},{"ecosystem":"PyPI","package":"vllm","purl":"pkg:pypi/vllm","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0.5.5"},{"fixed":"0.18.0"}]}],"versions":["0.10.0","0.10.1","0.10.1.1","0.10.2","0.11.0","0.11.1","0.11.2","0.12.0","0.13.0","0.14.0","0.14.1","0.15.0","0.15.1","0.16.0","0.17.0","0.17.1","0.5.5","0.6.0","0.6.1","0.6.1.post1","0.6.1.post2","0.6.2","0.6.3","0.6.3.post1","0.6.4","0.6.4.post1","0.6.5","0.6.6","0.6.6.post1","0.7.0","0.7.1","0.7.2","0.7.3","0.8.0","0.8.1","0.8.2","0.8.3","0.8.4","0.8.5","0.8.5.post1","0.9.0","0.9.0.1","0.9.1","0.9.2"],"fixed":["0.18.0"],"provider":"osv","source":"PYSEC-2026-2299","relation":"self"}],"product_status":[{"vendor":"Red Hat","product":"Red Hat AI Inference Server","cpe":"cpe:/a:redhat:ai_inference_server:3","status":"affected","packages":["rhaiis-preview/vllm-cuda-rhel9","rhaiis/vllm-cpu-rhel9","rhaiis/vllm-cuda-rhel9","rhaiis/vllm-neuron-rhel9","rhaiis/vllm-rocm-rhel9","rhaiis/vllm-spyre-rhel9","rhaiis/vllm-tpu-rhel9"],"package_count":7,"fixed":[],"advisories":[],"note":"Fix deferred","provider":"redhat-vex","source":"CVE-2026-34760"},{"vendor":"Red Hat","product":"Red Hat Enterprise Linux AI (RHEL AI) 3","cpe":"cpe:/a:redhat:enterprise_linux_ai:3","status":"affected","packages":["rhelai3/bootc-aws-cuda-rhel9","rhelai3/bootc-azure-cuda-rhel9","rhelai3/bootc-azure-rocm-rhel9","rhelai3/bootc-cuda-rhel9","rhelai3/bootc-gcp-cuda-rhel9","rhelai3/bootc-rocm-rhel9"],"package_count":6,"fixed":[],"advisories":[],"note":"Fix deferred","provider":"redhat-vex","source":"CVE-2026-34760"},{"vendor":"Red Hat","product":"Red Hat OpenShift AI (RHOAI)","cpe":"cpe:/a:redhat:openshift_ai","status":"affected","packages":["rhoai/odh-kserve-agent-rhel9","rhoai/odh-kserve-controller-rhel9","rhoai/odh-kserve-router-rhel9","rhoai/odh-kserve-storage-initializer-rhel9","rhoai/odh-vllm-cpu-rhel9","rhoai/odh-vllm-cuda-rhel9","rhoai/odh-vllm-gaudi-rhel9","rhoai/odh-vllm-rocm-rhel9"],"package_count":8,"fixed":[],"advisories":[],"note":"Fix deferred","provider":"redhat-vex","source":"CVE-2026-34760"}],"records":[{"provider":"csaf-bsi","source_id":"WID-SEC-W-2026-0987","relation":"upstream","url":"https://wid.cert-bund.de/portal/wid/securityadvisory?name=WID-SEC-2026-0987","license":"TLP:CLEAR, mirroring permitted","published_at":"2026-04-07T22:00:00Z","modified_at":"2026-08-19T22:00:00Z","summary":"vllm: Mehrere Schwachstellen","details":"Ein Angreifer kann mehrere Schwachstellen in vllm ausnutzen, um Dateien zu manipulieren, Sicherheitsmaßnahmen zu umgehen, vertrauliche Informationen offenzulegen oder einen Denial-of-Service-Zustand herbeizuführen.\n\nOpen Source vLLM ist eine Open-Source-Bibliothek für schnelle und effiziente Inferenz von Large Language Models (LLMs).","severity":[{"type":"BSI","score":"mittel","source":"WID-SEC-W-2026-0987"}],"affected":[],"product_status":[],"references":[{"type":"WEB","url":"https://wid.cert-bund.de/.well-known/csaf/white/2026/wid-sec-w-2026-0987.json"},{"type":"WEB","url":"https://wid.cert-bund.de/portal/wid/securityadvisory?name=WID-SEC-2026-0987"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-pf3h-qjgv-vcpr"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-pq5c-rjhq-qp7p"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-3mwp-wvh9-7528"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-6c4r-fmh3-7rh8"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/releases"},{"type":"WEB","url":"https://access.redhat.com/errata/RHSA-2026:36006"},{"type":"WEB","url":"https://access.redhat.com/errata/RHSA-2026:57389"},{"type":"WEB","url":"https://access.redhat.com/errata/RHSA-2026:57390"}]},{"provider":"cvelistv5","source_id":"CVE-2026-34760","relation":"self","url":"https://www.cve.org/CVERecord?id=CVE-2026-34760","license":"CC-BY-4.0","published_at":"2026-04-02T18:59:49.638Z","modified_at":"2026-04-03T14:42:34.842Z","summary":"vLLM: Downmix Implementation Differences as Attack Vectors Against Audio AI Models","details":"vLLM is an inference and serving engine for large language models (LLMs). From version 0.5.5 to before version 0.18.0, Librosa defaults to using numpy.mean for mono downmixing (to_mono), while the international standard ITU-R BS.775-4 specifies a weighted downmixing algorithm. This discrepancy results in inconsistency between audio heard by humans (e.g., through headphones/regular speakers) and audio processed by AI models (Which infra via Librosa, such as vllm, transformer). This issue has been patched in version 0.18.0.","severity":[{"type":"CVSS_V3","score":"CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:N/I:H/A:L","source":"CVE-2026-34760"}],"affected":[{"ecosystem":"CVE","package":"vllm-project/vllm","versions":[">= 0.5.5, < 0.18.0"],"provider":"cvelistv5","source":"CVE-2026-34760","relation":"self"}],"product_status":[],"references":[{"type":"WEB","url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-6c4r-fmh3-7rh8"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/pull/37058"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/commit/c7f98b4d0a63b32ed939e2b6dfaa8a626e9b46c4"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/releases/tag/v0.18.0"}]},{"provider":"nvd","source_id":"CVE-2026-34760","relation":"self","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-34760","severity":[],"affected":[{"ecosystem":"NVD","package":"vllm:vllm","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0.5.5"},{"fixed":"0.18.0"}]}],"fixed":["0.18.0"],"provider":"nvd","source":"CVE-2026-34760","relation":"self"}],"product_status":[],"references":[]},{"provider":"osv","source_id":"GHSA-6c4r-fmh3-7rh8","relation":"self","url":"https://osv.dev/vulnerability/GHSA-6c4r-fmh3-7rh8","license":"CC-BY-4.0","published_at":"2026-07-17T16:52:53Z","modified_at":"2026-07-17T17:00:52.155049Z","summary":"vLLM: Processing differential in multi-channel audio downmixing enables hidden-input/moderation bypass for audio models","details":"## Issue Description\nLibrosa defaults to using `numpy.mean` for mono downmixing (`to_mono`), while the international standard ITU-R BS.775-4 specifies a weighted downmixing algorithm. This discrepancy results in:\n- Inconsistency between audio heard by humans (e.g., through headphones/regular speakers) and audio processed by AI models (Which infra via Librosa, such as vllm, transformer).\n\nhttps://github.com/librosa/librosa/blob/af8c839fb15317fa2712ea66e7a22da6a9267b32/librosa/core/audio.py#L478\n## Attack Scenario and Impact\n\n### LFE (Low-Frequency Effects) Channel Exploit\nAttackers can craft special multichannel audio files containing:\n1. Normal content in front channels (L/R)\n2. Either interference signals or hidden content in the LFE channel\n\n**Notice:** It is worth noting that not only the LFE channel is excluded, but in fact, channels beyond the 6th (such as rear surround channels, overhead channels, height speakers, etc.) are also not supported.\n\n**Attack Methodology:**\n\nAttackers can create specially engineered multichannel audio with LFE interference, where front channels (L/R) contain normal content while the LFE channel carries interference signals or hidden content. When played on consumer devices that ignore LFE channels, only the normal content is heard. However, when processed by AI systems using Librosa (which mixes all channels), the LFE interference affects speech recognition feature extraction or masks critical detection features. This enables malicious content to bypass AI detection while still reaching end users, potentially compromising voice authentication systems, evading content moderation, or disrupting speech recognition accuracy.\n\n**Potential Exploitation Scenarios:**\n- Voice authentication systems may be tricked into accepting anomalous audio\n- Content moderation systems may fail to detect prohibited content hidden in LFE channels \n- Speech recognition systems may produce incorrect transcriptions\n\n**Note:** `torch.audio` implements this correctly. Failure to do so may lead to inconsistencies between training and test audio, resulting in performance degradation.\n\n\n## Resources\n\n- [ITU-R BS.775-4 Standard](https://www.itu.int/dms_pubrec/itu-r/rec/bs/R-REC-BS.775-4-202212-I!!PDF-E.pdf)\n- [Librosa Source Code](https://github.com/librosa/librosa/blob/af8c839fb15317fa2712ea66e7a22da6a9267b32/librosa/core/audio.py#L478)\n- [Librosa securty report](https://github.com/librosa/librosa/security/advisories/GHSA-vfm7-86xr-5mrh)\n\n## Fixes\n\n- https://github.com/vllm-project/vllm/pull/37058, which removes the librosa dependency from vLLM.","severity":[{"type":"CVSS_V3","score":"CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:N/I:H/A:L","source":"GHSA-6c4r-fmh3-7rh8"}],"affected":[{"ecosystem":"PyPI","package":"vllm","purl":"pkg:pypi/vllm","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0.5.5"},{"fixed":"0.18.0"}]}],"versions":["0.10.0","0.10.1","0.10.1.1","0.10.2","0.11.0","0.11.1","0.11.2","0.12.0","0.13.0","0.14.0","0.14.1","0.15.0","0.15.1","0.16.0","0.17.0","0.17.1","0.5.5","0.6.0","0.6.1","0.6.1.post1","0.6.1.post2","0.6.2","0.6.3","0.6.3.post1","0.6.4","0.6.4.post1","0.6.5","0.6.6","0.6.6.post1","0.7.0","0.7.1","0.7.2","0.7.3","0.8.0","0.8.1","0.8.2","0.8.3","0.8.4","0.8.5","0.8.5.post1","0.9.0","0.9.0.1","0.9.1","0.9.2"],"fixed":["0.18.0"],"provider":"osv","source":"GHSA-6c4r-fmh3-7rh8","relation":"self"}],"product_status":[],"references":[{"type":"WEB","url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-6c4r-fmh3-7rh8"},{"type":"ADVISORY","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-34760"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/pull/37058"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/commit/c7f98b4d0a63b32ed939e2b6dfaa8a626e9b46c4"},{"type":"WEB","url":"https://github.com/pypa/advisory-database/tree/main/vulns/vllm/PYSEC-2026-2299.yaml"},{"type":"PACKAGE","url":"https://github.com/vllm-project/vllm"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/releases/tag/v0.18.0"}]},{"provider":"osv","source_id":"PYSEC-2026-2299","relation":"self","url":"https://osv.dev/vulnerability/PYSEC-2026-2299","license":"CC-BY-4.0","published_at":"2026-04-02T20:16:25.437Z","modified_at":"2026-07-13T07:15:12.553209Z","details":"vLLM is an inference and serving engine for large language models (LLMs). From version 0.5.5 to before version 0.18.0, Librosa defaults to using numpy.mean for mono downmixing (to_mono), while the international standard ITU-R BS.775-4 specifies a weighted downmixing algorithm. This discrepancy results in inconsistency between audio heard by humans (e.g., through headphones/regular speakers) and audio processed by AI models (Which infra via Librosa, such as vllm, transformer). This issue has been patched in version 0.18.0.","severity":[{"type":"CVSS_V3","score":"CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:L","source":"PYSEC-2026-2299"}],"affected":[{"ecosystem":"PyPI","package":"vllm","purl":"pkg:pypi/vllm","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0.5.5"},{"fixed":"0.18.0"}]}],"versions":["0.10.0","0.10.1","0.10.1.1","0.10.2","0.11.0","0.11.1","0.11.2","0.12.0","0.13.0","0.14.0","0.14.1","0.15.0","0.15.1","0.16.0","0.17.0","0.17.1","0.5.5","0.6.0","0.6.1","0.6.1.post1","0.6.1.post2","0.6.2","0.6.3","0.6.3.post1","0.6.4","0.6.4.post1","0.6.5","0.6.6","0.6.6.post1","0.7.0","0.7.1","0.7.2","0.7.3","0.8.0","0.8.1","0.8.2","0.8.3","0.8.4","0.8.5","0.8.5.post1","0.9.0","0.9.0.1","0.9.1","0.9.2"],"fixed":["0.18.0"],"provider":"osv","source":"PYSEC-2026-2299","relation":"self"}],"product_status":[],"references":[{"type":"ADVISORY","url":"https://github.com/vllm-project/vllm/releases/tag/v0.18.0"},{"type":"ADVISORY","url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-6c4r-fmh3-7rh8"},{"type":"REPORT","url":"https://github.com/vllm-project/vllm/pull/37058"},{"type":"FIX","url":"https://github.com/vllm-project/vllm/commit/c7f98b4d0a63b32ed939e2b6dfaa8a626e9b46c4"}]},{"provider":"redhat-vex","source_id":"CVE-2026-34760","relation":"self","url":"https://access.redhat.com/security/cve/CVE-2026-34760","license":"CC-BY-4.0","published_at":"2026-04-02T18:59:49.638Z","modified_at":"2026-05-11T13:30:06Z","summary":"vLLM: Librosa: numpy: Librosa: AI model data integrity impact due to audio processing discrepancy","details":"A flaw was found in Librosa, a software library used by artificial intelligence (AI) models like vLLM for processing audio. The library's method for converting stereo audio to mono differs from international standards, causing AI models to interpret audio differently than humans. This inconsistency could allow an attacker to provide specially crafted audio, leading to the AI model processing the data incorrectly. This could result in a high impact on the integrity of the AI system's operations.","severity":[{"type":"CVSS_V3","score":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:H/A:N","source":"CVE-2026-34760"},{"type":"Red Hat","score":"Moderate","source":"CVE-2026-34760"}],"affected":[],"product_status":[{"vendor":"Red Hat","product":"Red Hat AI Inference Server","cpe":"cpe:/a:redhat:ai_inference_server:3","status":"affected","packages":["rhaiis-preview/vllm-cuda-rhel9","rhaiis/vllm-cpu-rhel9","rhaiis/vllm-cuda-rhel9","rhaiis/vllm-neuron-rhel9","rhaiis/vllm-rocm-rhel9","rhaiis/vllm-spyre-rhel9","rhaiis/vllm-tpu-rhel9"],"package_count":7,"fixed":[],"advisories":[],"note":"Fix deferred","provider":"redhat-vex","source":"CVE-2026-34760"},{"vendor":"Red Hat","product":"Red Hat Enterprise Linux AI (RHEL AI) 3","cpe":"cpe:/a:redhat:enterprise_linux_ai:3","status":"affected","packages":["rhelai3/bootc-aws-cuda-rhel9","rhelai3/bootc-azure-cuda-rhel9","rhelai3/bootc-azure-rocm-rhel9","rhelai3/bootc-cuda-rhel9","rhelai3/bootc-gcp-cuda-rhel9","rhelai3/bootc-rocm-rhel9"],"package_count":6,"fixed":[],"advisories":[],"note":"Fix deferred","provider":"redhat-vex","source":"CVE-2026-34760"},{"vendor":"Red Hat","product":"Red Hat OpenShift AI (RHOAI)","cpe":"cpe:/a:redhat:openshift_ai","status":"affected","packages":["rhoai/odh-kserve-agent-rhel9","rhoai/odh-kserve-controller-rhel9","rhoai/odh-kserve-router-rhel9","rhoai/odh-kserve-storage-initializer-rhel9","rhoai/odh-vllm-cpu-rhel9","rhoai/odh-vllm-cuda-rhel9","rhoai/odh-vllm-gaudi-rhel9","rhoai/odh-vllm-rocm-rhel9"],"package_count":8,"fixed":[],"advisories":[],"note":"Fix deferred","provider":"redhat-vex","source":"CVE-2026-34760"}],"references":[{"type":"WEB","url":"https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-34760.json"},{"type":"WEB","url":"https://access.redhat.com/security/cve/CVE-2026-34760"},{"type":"WEB","url":"https://bugzilla.redhat.com/show_bug.cgi?id=2454645"},{"type":"WEB","url":"https://www.cve.org/CVERecord?id=CVE-2026-34760"},{"type":"WEB","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-34760"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/commit/c7f98b4d0a63b32ed939e2b6dfaa8a626e9b46c4"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/pull/37058"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/releases/tag/v0.18.0"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-6c4r-fmh3-7rh8"}]}],"references":[{"type":"WEB","url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-6c4r-fmh3-7rh8"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/pull/37058"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/commit/c7f98b4d0a63b32ed939e2b6dfaa8a626e9b46c4"},{"type":"WEB","url":"https://github.com/vllm-project/vllm/releases/tag/v0.18.0"},{"type":"ADVISORY","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-34760"},{"type":"WEB","url":"https://github.com/pypa/advisory-database/tree/main/vulns/vllm/PYSEC-2026-2299.yaml"},{"type":"PACKAGE","url":"https://github.com/vllm-project/vllm"},{"type":"WEB","url":"https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-34760.json"},{"type":"WEB","url":"https://access.redhat.com/security/cve/CVE-2026-34760"},{"type":"WEB","url":"https://bugzilla.redhat.com/show_bug.cgi?id=2454645"},{"type":"WEB","url":"https://www.cve.org/CVERecord?id=CVE-2026-34760"}],"sources":[{"provider":"csaf-bsi","source_id":"WID-SEC-W-2026-0987","url":"https://wid.cert-bund.de/portal/wid/securityadvisory?name=WID-SEC-2026-0987","license":"TLP:CLEAR, mirroring permitted","modified_at":"2026-08-19T22:00:00Z","relation":"upstream"},{"provider":"cvelistv5","source_id":"CVE-2026-34760","url":"https://www.cve.org/CVERecord?id=CVE-2026-34760","license":"CC-BY-4.0","modified_at":"2026-04-03T14:42:34.842Z","relation":"self"},{"provider":"nvd","source_id":"CVE-2026-34760","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-34760","relation":"self"},{"provider":"osv","source_id":"GHSA-6c4r-fmh3-7rh8","url":"https://osv.dev/vulnerability/GHSA-6c4r-fmh3-7rh8","license":"CC-BY-4.0","modified_at":"2026-07-17T17:00:52.155049Z","relation":"self"},{"provider":"osv","source_id":"PYSEC-2026-2299","url":"https://osv.dev/vulnerability/PYSEC-2026-2299","license":"CC-BY-4.0","modified_at":"2026-07-13T07:15:12.553209Z","relation":"self"},{"provider":"redhat-vex","source_id":"CVE-2026-34760","url":"https://access.redhat.com/security/cve/CVE-2026-34760","license":"CC-BY-4.0","modified_at":"2026-05-11T13:30:06Z","relation":"self"}]}
