MediumVulnerabilityLLM-specific
CVE-2026-34760: vLLM audio mono downmixing mismatch with ITU-R BS.775-4 standard
- Source
- NVD(opens in a new tab)
- Identifier
- CVE-2026-34760
- Published
- Record updated
- Affected
- vllm >= 0.5.5, < 0.18.0
- Fixed in
- 0.18.0
- released March 20, 2026, 13 days before disclosure
- Known exploitation
- Not listed in the CISA Known Exploited Vulnerabilities catalog at the last check.
- EPSS
- 0.3%
Summary
vLLM, an inference and serving engine for large language models, versions 0.5.5 to before 0.18.0, uses Librosa's default numpy.mean for mono downmixing (to_mono). The ITU-R BS.775-4 standard specifies a weighted downmix, so audio heard by humans differs from audio processed by AI models such as those served by vLLM or Transformers.
Mitigation
This issue has been patched in version 0.18.0.
Affected packages in the Exposure Registry
Matched by package name and ecosystem. Each entry shows whether the package delegates to a language model and how many tracked packages depend on it.
- vllmPyPILLM dependency since 2023-08-25 · 15 tracked dependents
Topics
Related items
- CriticalGHSA-9x8x-crjq-g64h: Jupyter MCP Server: Browser DNS rebinding can hijack the Jupyter backend through management routesSimilar attack · GitHub Advisory Database
- HighGHSA-24pw-rf5p-mgp7: Jupyter MCP Server: Unauthenticated management routes can hijack the Jupyter backend in standalone HTTP modeSimilar attack · GitHub Advisory Database
- HighGHSA-m932-63jc-4jqq: k8s-mcp-server: Missing authentication on default SSE/HTTP transport allows unauthenticated Kubernetes tool executionSimilar attack · GitHub Advisory Database
- HighGHSA-3mgh-93r5-r7jf: CodeWhale: exec_shell parallel-readonly classifier auto-approves fd/rg/git grep exec flags enabling silent RCESimilar attack · GitHub Advisory Database
- HighGHSA-qx49-fqc8-xw99: Model Context Protocol Python SDK: OAuth client could send credentials to an authorization server chosen by the MCP serverSimilar attack · GitHub Advisory Database