CVE-2026-105752: vllm-project vllm

Vulnerability summary

CVE-2026-105752 is a low (CVSS 3.1) vulnerability in vllm-project vllm. EPSS estimates a 0.22% chance of exploitation in the next 30 days, higher than 11.4% of scored vulnerabilities.

Timeline

1. 05-10-2026
   PublishedVulnerability first disclosed.
2. 06-10-2026
   Last modifiedVulnerability information updated.

Description

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, Harmony tool continuations submitted through "POST /v1/responses" requests rebuild the next-turn engine input without preserving the cache_salt value, placing the continuation prefix in the global unsalted cache namespace even when the caller enabled salting. On deployments with prefix caching enabled, which is the default, an authenticated tenant who can reconstruct a victim's low-entropy post-tool history can submit the same continuation and use the cached_tokens_per_turn count to determine whether the prefix was previously processed, defeating the intended tenant isolation of salted prefix caching. This issue is fixed in version 0.30.0.

• CWE-200Exposure of Sensitive Information to an Unauthorized ActorThe product exposes sensitive information to an actor that is not explicitly authorized to have access to that information.
• CWE-524Use of Cache Containing Sensitive InformationThe code uses a cache that contains sensitive information, but the cache can be read by an actor outside of the intended control sphere.