vllm-project · vLLM · published 2026-04-02

CVE-2026-34760vLLM: Downmix Implementation Differences as Attack Vectors Against Audio AI Models

Routine No exploitation evidence; EPSS 0.3%.

CVE-2026-34760 is not known to be exploited in the wild. EPSS estimates a 0.3% chance of exploitation in the next 30 days, higher than 21% of all CVEs. This vLLM vulnerability: CVSS 5.9 (v3.1). Fixed upstream in vLLM 0.18.0. Routine: patch when convenient.

Share this

Remediation#

Fixed upstream in vLLM 0.18.0.

Patch events

  1. This monthSecurity release

    AI toolingvLLM

    vLLM security release: 0.18.0

    Upgrade to 0.18.0

    Why: Vendor security release for server software (most severe: CVE-2026-34760).

Exploit intelligence#

SignalValueSource
Known exploitedNoCISA KEV
Ransomware use—CISA KEV
EPSS (30-day exploitation probability)0.3% (21th pct)FIRST EPSS v2026.06.15
CVSS base score5.9 (v3.1)
CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:N/I:H/A:L
CNA / CISA ADP
SSVC exploitationnoneCISA Vulnrichment
SSVC automatableNoCISA Vulnrichment
SSVC technical impactpartialCISA Vulnrichment

Affected versions (from the CVE record)#

Description (from the CVE record)#

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.

Quoted from the CVE Program record.

Elsewhere#

Distro status: Debian · Ubuntu · Red Hat · Alpine

Something wrong on this page?
Anonymous. Reports are checked against the primary source.

Also on today's list

See today's list → · Coming up →