vLLM is an inference and serving engine for large language models. Prior to 0.30.0, flash late-interaction scoring at the /score and /rerank endpoints derives each worker's query_key value from the caller-controlled X-Request-Id header. A concurrent request that reuses a victim's identifier can overwrite the cached query embedding so the victim's documents are scored against the attacker's query, and shared use counters can also cause a late-interaction cache-miss error. This issue is fixed in version 0.30.0.

There are packages installed that are affected by a vulnerability referenced in the following CVE:

  - vLLM is an inference and serving engine for large language models. Prior to 0.30.0, flash late-interaction
    scoring at the /score and /rerank endpoints derives each worker's query_key value from the caller-
    controlled X-Request-Id header. A concurrent request that reuses a victim's identifier can overwrite the
    cached query embedding so the victim's documents are scored against the attacker's query, and shared use
    counters can also cause a late-interaction cache-miss error. This issue is fixed in version 0.30.0.
    (CVE-2026-105755)

ID	Name	Product	Family	Type	Severity
476117	Chainguard: vllm-openai-cuda-13.0: security update to 0.31.0-r0	Tenable Cloud Security	Chainguard Local Security Checks	Local	medium
476117	Chainguard: vllm-openai-cuda-13.0: security update to 0.31.0-r0	Tenable Self-Hosted Container Security	Chainguard Local Security Checks	Local	medium
473004	SCA: security update for vllm (GHSA-2phq-3phc-84px)	Tenable Cloud Security	SCA Checks	Local	medium
473004	SCA: security update for vllm (GHSA-2phq-3phc-84px)	Tenable Self-Hosted Container Security	SCA Checks	Local	medium

Detections

Analytics

CVE-2026-105755

medium

Tenable Plugins

medium

medium

medium

medium