THREATOPS
THREAT OPS › Threat News › [NVD] CVE-2026-105755 (MEDIUM 4.2) — 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

[NVD] CVE-2026-105755 (MEDIUM 4.2) — 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

lownvdPublished 2026-10-05

CVE-2026-105755 CVSS: 4.2 MEDIUM Published: 2026-10-05T23:17:02.167

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 embeddin

Indicators of compromise

Original source: https://nvd.nist.gov/vuln/detail/CVE-2026-105755