vLLM <= 0.23.0 - Anthropic Router Heap Address Information Leak

vLLM = 0.23.0 incompletely fixes CVE-2026-22778. The original fix added sanitizemessage to the OpenAI router but the Anthropic-compatible router /v1/messages echoes strexc directly. id: CVE-2026-54236 info: name: vLLM = 0.23.0 - Anthropic Router Heap Address Information Leak author: kenlacroix...

EXPLOIT-CVE-2026-22778

CVE-2026-22778 — vLLM 视频处理中的 RCE 针对 CVE-2026-22778 （CVSS 9.8）的漏洞实验环境与概念验证，该漏洞是 vLLM 多模态摄取路径中的未认证远程代码执行链。 | ---|--- CVE| CVE-2026-22778 公告| GHSA-4r2x-xpjr-7cvv 受影响版本| vLLM = 0.8.3, vLLM 将媒体加载失败转换为 HTTP 400，并将 exc.detail 原封不动地返回给客户端 （apiserver.py）： async def httpexceptionhandler: Request, exc:...

CVE-2026-22807_Range

🛡️ CVE-2026-22807 Simulation: AI Supply Chain RCE Target Range ⚠️ 免责声明 / Disclaimer 本项目仅用于安全研究与教学目的 ，旨在演示 AI 模型加载过程中的供应链攻击原理（Remote Code Execution）。 请勿在生产环境中使用生成的恶意模型，也严禁将本项目用于任何非法攻击行为。开发者不承担任何因使用不当造成的法律责任。 📖 项目简介 Introduction 这是一个轻量级的 Python 靶场，用于模拟和复现 CVE-2026-22807 （一个典型的 AI 供应链漏洞逻辑）。 该靶场演示了当 AI...

LeakGauge

LeakGauge 这是我们论文的代码仓库：The Model's Tell: Measuring Context-Leakage Attack Signals with Behavior Gauges。 ArXiv 版本和论文链接：https://arxiv.org/abs/2608.17829 本仓库实现了 LeakGauge 泄漏检测流水线：演示数据集的准备、log-probability 提取、探针训练以及在线或离线检测。 关于其他安全与防护任务，请参阅 SafeGauge。 引用 @misczhang2026leakgauge, title=The Model's Tell:...

vLLM 0.8.3 < 0.14.1 Information Disclosure (GHSA-4r2x-xpjr-7cvv)

The version of the vLLM Python package installed on the remote host is 0.8.3 prior to 0.14.1. It is, therefore, affected by an information disclosure vulnerability. When an invalid image is sent to the vLLM multimodal endpoint, PIL throws an error that vLLM returns to the client, leaking a heap...

CVE-2026-105922

A flaw was found in vLLM. A remote attacker can exploit this vulnerability by sending manipulated input processed by the penalty handling mechanism. This triggers an unhandled failure in the token bin counting routine gettokenbincountsandmask, crashing the service and resulting in a Denial of...

CVE-2026-105922

A security flaw has been discovered in vllm-project vLLM up to 0.31.0. This impacts the function gettokenbincountsandmask of the file vllm/modelexecutor/layers/utils.py of the component Penalty Handler. Performing a manipulation results in denial of service. Remote exploitation of the attack is...

CVE-2026-105922

A security flaw exists in the vllm-project vLLM software (up to version 0.31.0 ) affecting the Penalty Handler component. The vulnerability resides in the get_token_bin_counts_and_mask function within the file vllm/model_executor/layers/utils.py. An attacker can perform a manipulation resulting i...

EUVD-2026-93430

A security flaw has been discovered in vllm-project vLLM up to 0.31.0. This impacts the function gettokenbincountsandmask of the file vllm/modelexecutor/layers/utils.py of the component Penalty Handler. Performing a manipulation results in denial of service. Remote exploitation of the attack is...

CVE-2026-105922 vllm-project vLLM Penalty utils.py get_token_bin_counts_and_mask denial of service

A security flaw has been discovered in vllm-project vLLM up to 0.31.0. This impacts the function gettokenbincountsandmask of the file vllm/modelexecutor/layers/utils.py of the component Penalty Handler. Performing a manipulation results in denial of service. Remote exploitation of the attack is...

CVE-2026-105922 vllm-project vLLM Penalty utils.py get_token_bin_counts_and_mask denial of service

A security flaw has been discovered in vllm-project vLLM up to 0.31.0. This impacts the function gettokenbincountsandmask of the file vllm/modelexecutor/layers/utils.py of the component Penalty Handler. Performing a manipulation results in denial of service. Remote exploitation of the attack is...

CVE-2026-105775

A flaw was found in vLLM. An authenticated remote attacker can send a crafted request that triggers an out-of-bounds read in the completions request handler. This issue can cause the service to crash, resulting in a Denial of Service DoS...

CVE-2026-105775

A security vulnerability has been detected in vllm-project vLLM up to 0.31.0. This impacts the function convssmforward of the file vllm/modelexecutor/layers/mamba/mambamixer2.py of the component Completions Request Handler. The manipulation leads to out-of-bounds read. The attack is possible to b...

CVE-2026-105775 vllm-project vLLM Completions Request mamba_mixer2.py conv_ssm_forward out-of-bounds

A security vulnerability has been detected in vllm-project vLLM up to 0.31.0. This impacts the function convssmforward of the file vllm/modelexecutor/layers/mamba/mambamixer2.py of the component Completions Request Handler. The manipulation leads to out-of-bounds read. The attack is possible to b...

CVE-2026-105775 vllm-project vLLM Completions Request mamba_mixer2.py conv_ssm_forward out-of-bounds

A security vulnerability has been detected in vllm-project vLLM up to 0.31.0. This impacts the function convssmforward of the file vllm/modelexecutor/layers/mamba/mambamixer2.py of the component Completions Request Handler. The manipulation leads to out-of-bounds read. The attack is possible to b...

EUVD-2026-92962

A security vulnerability has been detected in vllm-project vLLM up to 0.31.0. This impacts the function convssmforward of the file vllm/modelexecutor/layers/mamba/mambamixer2.py of the component Completions Request Handler. The manipulation leads to out-of-bounds read. The attack is possible to b...

CVE-2026-105775

A security vulnerability exists in vllm-project vLLM up to version 0.31.0 . The issue resides within the Completions Request Handler component, specifically affecting the conv_ssm_forward function in the file vllm/model_executor/layers/mamba/mamba_mixer2.py . This flaw allows for an out-of-bounds...

CVE-2026-105752

A flaw was found in vLLM. When multi-turn requests using tool continuations are processed, the service fails to retain the cache salt parameter, causing cached data to be placed in a shared, unsalted cache namespace rather than an isolated tenant cache. An authenticated tenant with knowledge of a...

CVE-2026-105760

A flaw was found in vLLM. A remote attacker can trigger a Denial of Service DoS by submitting a request with excessively large frame rate and frame count options for the GLMGA video backend. Because the application constructs intermediate frame index lists without proper upper bounds, processing...

CVE-2026-105759

A flaw was found in vLLM. An unauthenticated remote attacker can cause a denial of service DoS by sending requests with arbitrary, non-standard HTTP method names to unguarded server routes. Because the metrics tracking middleware records raw request methods into monitoring labels without validati...