CVE-2026-34760 is a medium-severity (CVSS 5.9) Improper Input Validation vulnerability in vllm. A fix is available for vllm — see the affected versions and patch details below.
vLLM: Downmix Implementation Differences as Attack Vectors Against Audio AI Models
Exploitation Status
No confirmed exploitation observed yet
- CISA’s own triage has not observed active exploitation or public proof-of-concept code for this CVE as of its last assessment.
Exploitation and automatability from CISA’s SSVC triage for CVE-2026-34760.
EPSS Exploitation Probability
Probability of exploitation in the next 30 days, from FIRST.org EPSS.
How urgent is this, really
CVE-2026-34760 by exploitation likelihood (EPSS) against impact (CVSS). Outside the shaded patch-first corner.
Where this sits among everything scored
Of 395,249 CVEs with a current EPSS score, this one falls in the < 10% band (highlighted). Counts from FIRST.org, log-scaled.
Real-World Exposure
vllmReal-time download stats are indexed for npm and PyPI packages. This vulnerability affects PyPI packages — download data is not available via public APIs for these ecosystems.
Description
Issue Description
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).
Attack Scenario and Impact
LFE (Low-Frequency Effects) Channel Exploit
Attackers can craft special multichannel audio files containing:
- Normal content in front channels (L/R)
- Either interference signals or hidden content in the LFE channel
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.
Attack Methodology:
Attackers 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.
Potential Exploitation Scenarios:
- Voice authentication systems may be tricked into accepting anomalous audio
- Content moderation systems may fail to detect prohibited content hidden in LFE channels
- Speech recognition systems may produce incorrect transcriptions
Note: torch.audio implements this correctly. Failure to do so may lead to inconsistencies between training and test audio, resulting in performance degradation.
Resources
Fixes
- https://github.com/vllm-project/vllm/pull/37058, which removes the librosa dependency from vLLM.
Affected Packages
| Ecosystem | Package | Vulnerable range | Fix |
|---|---|---|---|
| 🐍PyPI | vllm | ≥ 0.5.5&&< 0.18.0 | 0.18.0pip install --upgrade 'vllm==0.18.0' |
Affected Products
vllmvllmDetection & mitigation playbook
Open-source dependencyDetect
Scan your dependency tree (package-lock.json, pnpm-lock.yaml, requirements.txt, go.sum, etc.) for vllm, including transitive dependencies — a direct dependency you never call can still pull in a vulnerable version.
Fix
Update vllm to 0.18.0 or later, then make sure no transitive (indirect) dependency still pins the vulnerable range — O3 confirms CVE-2026-34760 is resolved across your whole dependency graph.
Workarounds
If you can't upgrade right away: gate or disable the affected feature, validate untrusted input at the boundary, and avoid passing attacker-controlled data into the vulnerable path. O3's runtime protection blocks exploitation in production as an interim safeguard until the upgrade lands.
Fixing This On Your OS
If you run this on a Linux distribution, patch through your package manager against the distro's own security advisory below — it tracks the exact backported fix for your release, which can ship on a different timeline (and sometimes a different severity) than the upstream project.
The vLLM inference and serving engine for LLM models uses a flawed version of librosa, which downmix audios to mono while international standards mandates it should be used a weight downmixing algorithm. When vLLM is processing an audio file this may result in discrepancy between what's processed by the AI model being…
Mitigation for this issue is either not available or the currently available options do not meet the Red Hat Product Security criteria comprising ease of use and deployment, applicability to widespread installation base, or stability.Source: Red Hat security advisory for CVE-2026-34760 (CC BY 4.0)
Frequently Asked Questions
Is CVE-2026-34760 in your dependencies?
Find it across PyPI, including transitive dependencies.