Ghostwire

CVE-2026-54234: vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Prior to 0.24.0, a frontend-legal...

HIGH CVSS 0.0

Published: July 6, 2026 | Last Modified: July 6, 2026

Description

vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Prior to 0.24.0, a frontend-legal multi-request speculative decoding workload can cause the rejection sampler to produce a recovered token equal to the model vocabulary size boundary value, which is then converted to negative one when the engine selects the next live token for a request and is written back into the drafter's input ids; that out-of-vocabulary value is later consumed by the model's embedding and attention path and crashes the engine worker with a GPU device-side assertion. The same triggering request sequence is reachable through the public gRPC Generate and Abort endpoints, so a remote client that can send generation requests can crash the shared engine worker, aborting concurrent requests and causing a service-wide denial of service for other clients of the deployment until the worker is restarted. This issue is fixed in version 0.24.0.

Ghostwire Analysis — What This Means Practically

This analysis is generated by Ghostwire from NVD, CISA KEV, EPSS, and open-source intelligence data. Verify findings through primary sources before acting.

Security Coverage (1 articles)

References