Goal Reached Thanks to every supporter — we hit 100%!

Goal: 1000 CNY · Raised: 1359 CNY

100%
Buy Us a Coffee

CVE-2026-54234— vLLM: Remote DoS in vLLM via Invalid Recovered Token Reinjection

Quick assessment

Affected
vllm-project vllm
Exploitation
Public or AI PoC available; prioritize validation
Recommended action
Check the vendor advisory and references for a fixed version. If immediate upgrade is impossible, restrict exposure and increase monitoring.

vLLM是vLLM团队开源的一个适用于 LLM 的高吞吐量和内存高效推理和服务引擎。 vLLM 0.24.0之前版本存在输入验证错误漏洞，该漏洞源于输入验证错误，可能导致远程客户端通过公开的gRPC Generate和Abort端点发送生成请求，造成引擎工作进程崩溃，从而引发服务范围内的拒绝服务。

CVSS 7.5 · High EPSS 0.62% · P48

Affected Version Matrix 1

VendorProduct Version RangeStatus
vllm-project vllm < 0.24.0 affected
Get alerts for future matching vulnerabilities Log in to subscribe

I. Basic Information for CVE-2026-54234

Vulnerability Information
Have questions about the vulnerability? See if Shenlong's analysis helps!
View Shenlong Deep Dive ↗

Although we use advanced large model technology, its output may still contain inaccurate or outdated information.Shenlong tries to ensure data accuracy, but please verify and judge based on the actual situation.

Vulnerability Title
vLLM: Remote DoS in vLLM via Invalid Recovered Token Reinjection
Source: CVE Program / CVE List V5
Vulnerability 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.
Source: CVE Program / CVE List V5
CVSS Information
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
Source: CVE Program / CVE List V5
Vulnerability Type
输入验证不恰当
Source: CVE Program / CVE List V5
Vulnerability Title
vLLM 输入验证错误漏洞
Source: CNNVD (China National Vulnerability Database)
Vulnerability Description
vLLM是vLLM团队开源的一个适用于 LLM 的高吞吐量和内存高效推理和服务引擎。 vLLM 0.24.0之前版本存在输入验证错误漏洞，该漏洞源于输入验证错误，可能导致远程客户端通过公开的gRPC Generate和Abort端点发送生成请求，造成引擎工作进程崩溃，从而引发服务范围内的拒绝服务。
Source: CNNVD (China National Vulnerability Database)
CVSS Information
N/A
Source: CNNVD (China National Vulnerability Database)
Vulnerability Type
N/A
Source: CNNVD (China National Vulnerability Database)

Affected Products

Vendor Product Affected Versions CPE Subscribe
vllm-project vllm < 0.24.0 -

II. Public POCs for CVE-2026-54234

# POC Description Source Link Shenlong Link
AI-Generated POC Premium
Qwen3.6-35B-A3B · 8098 chars
Pro+ exclusive includes:
Vulnerability reproduction recording (real sandbox build + trigger, exclusive)
In-depth vulnerability mechanism
Trigger conditions & impact
Full executable POC code
Exploit chain & mitigation
POC zip download
100+ AI POC generations per month

III. Intelligence Information for CVE-2026-54234

请登录查看更多情报信息。

Patches & Fixes for CVE-2026-54234 (2)

Vendor Advisories for CVE-2026-54234 (1)

Same Patch Batch · vllm-project · 2026-07-06 · 4 CVEs total

CVE-2026-55646 6.5 MEDIUM vLLM speech-to-text endpoints allocate full upload before enforcing the audio file-size li
CVE-2026-55514 vLLM denial of service via prompt embeds on M-RoPE models
CVE-2026-55574 vLLM: ReDoS via structured_outputs.regex compiled without timeout in xgrammar and outlines

IV. Related Vulnerabilities

Same product: vllm
Same vendor: vllm-project
Same weakness: CWE-20

V. Comments for CVE-2026-54234

No comments yet

Leave a comment