CVE-2026-105754
Description
vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out path accepts caller-supplied tensors in the features.kwargs_data field, cache identifiers in the features.mm_hashes field, ranges in the features.mm_placeholders field, and wire-selected multimodal field processors without rebinding them to the active model renderer contract. Forged grid geometry, field types, or non-positive placeholder lengths can terminate the shared EngineCore; when an attacker knows or can induce a victim's content hash, forged cache hashes can poison or retrieve cross-request encoder-cache state; and dropped sparse placeholder masks can alter replayed transport semantics. This issue is fixed in version 0.30.0.
AI Insight
LLM-synthesized narrative grounded in this CVE's description and references.
Affected packages
Versions sourced from the GitHub Security Advisory.
| Package | Affected versions | Patched versions |
|---|---|---|
vllmPyPI | < 0.30.0 | 0.30.0 |
Affected products
2Patches
Vulnerability mechanics
References
5- github.com/vllm-project/vllm/commit/1970f3ed4be7fa8620e4ddc4a12c36a8384cfc27nvdPatchWEB
- github.com/vllm-project/vllm/pull/51898nvdIssue TrackingPatchWEB
- github.com/advisories/GHSA-ph72-cqr5-qpp7ghsaADVISORY
- github.com/vllm-project/vllm/security/advisories/GHSA-ph72-cqr5-qpp7nvdMitigationVendor AdvisoryWEB
- github.com/vllm-project/vllm/releases/tag/v0.30.0nvdRelease NotesWEB
News mentions
1- VLLM: Batch of 11 Medium-Severity Vulnerabilities Disclosed TogetherVypr Intelligence · Oct 6, 2026