vllm-project · vLLM · published 2026-07-06
CVE-2026-54234vLLM: Remote DoS in vLLM via Invalid Recovered Token Reinjection
This month Public proof-of-concept exists (SSVC exploitation: poc).
CVE-2026-54234 has a public proof-of-concept exploit, but no confirmed exploitation in the wild. EPSS estimates a 0.6% chance of exploitation in the next 30 days, higher than 50% of all CVEs. This vLLM vulnerability: CVSS 7.5 (v3.1), attacks can be automated. Fixed upstream in vLLM 0.24.0. Patch in your normal monthly cycle.
Jump to: Remediation Exploit intelAffected versionsDescription
Remediation#
Fixed upstream in vLLM 0.24.0.
Patch events
AI toolingvLLM
vLLM security release: 0.24.0
Upgrade to
0.24.0Why: CVE-2026-54234: Public proof-of-concept exists (SSVC exploitation: poc).
Exploit intelligence#
| Signal | Value | Source |
|---|---|---|
| Known exploited | No | CISA KEV |
| Ransomware use | — | CISA KEV |
| EPSS (30-day exploitation probability) | 0.6% (50th pct) | FIRST EPSS v2026.06.15 |
| CVSS base score | 7.5 (v3.1)CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H | CNA / CISA ADP |
| SSVC exploitation | poc | CISA Vulnrichment |
| SSVC automatable | Yes | CISA Vulnrichment |
| SSVC technical impact | partial | CISA Vulnrichment |
Affected versions (from the CVE record)#
- vllm-project vllm
< 0.24.0(affected)
Description (from the CVE record)#
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.
Quoted from the CVE Program record.