Vulnerability · Published Oct 5, 2026 · Assigned by GitHub_M

CVE-2026-105755vLLM: Flash late-interaction scoring caches query embeddings under a caller-controlled request id — cross-request integrity break and induced errors on `/score` and `/rerank`

Medium 4.2

In plain English

A medium-severity authorization bypass via user-controlled key in vllm from vllm-project. There are no confirmed reports of exploitation so far. The chance it is exploited in the next 30 days is estimated at 0.2% (higher than 10% of all known vulnerabilities). Fixed versions are available - upgrade the affected package.

Severity (CVSS 3.1)4.2Medium
Exploit odds (30 days)0.2%top 89.7% of all CVEs
Exploited in the wildNo reportsCISA SSVC: none
Attack toolingNoneno public tooling known

Official description

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.

Affected packages & fixed versions

EcosystemPackageFixed inAdvisory
PyPIvllm0.30.0GHSA-2phq-3phc-84px
PyPIvllm0.30.0PYSEC-2026-4198

Affected products (as reported)

VendorProductVersions
vllm-project vllm < 030.0

Weakness: CWE-639 · Vector: CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:N/I:L/A:L

References