2534 matches found
CVE-2026-126394
The vLLM library (versions 0.30.0 before 0.31.0 ) contains a resource exhaustion vulnerability in the OpenCVDynamicOpenPanguVideoBackend . The issue stems from the component using request-controlled num_frames without an effective ceiling. Unauthenticated attackers can trigger a denial of service...
EUVD-2026-125171
vLLM 0.30.0 before 0.31.0 contains a resource exhaustion vulnerability in OpenCVDynamicOpenPanguVideoBackend, which uses request-controlled numframes without an effective ceiling. Unauthenticated attackers can send /v1/chat/completions or /tokenize requests selecting the openpangu videobackend wi...
CVE-2026-126395
The vLLM library (versions 0.14.0rc1 before 0.31.0) is vulnerable to a heap address disclosure in the sanitize_message() function. The root cause is a lowercase-only regex that fails to scrub uppercase PIL image addresses. Unauthenticated attackers can exploit this by sending a crafted PNG to end...
EUVD-2026-125170
vLLM from 0.14.0rc1 before 0.31.0 contains an information disclosure vulnerability in sanitizemessage whose lowercase-only regex fails to scrub uppercase PIL image addresses. Unauthenticated attackers can send a crafted PNG to endpoints such as /tokenize, triggering HF-processor errors that leak...
CVE-2026-126397
The vLLM library versions 0.10.1 through 0.31.0 are vulnerable to a Denial of Service (DoS) attack affecting Cohere2 Vision models. The root cause is the lack of a server-side upper bound on the mm_processor_kwargs.max_patches value, which is forwarded directly from request-controlled inputs to t...
CVE-2026-126396
The vLLM library (versions 0.24.0 to before 0.31.0 ) is vulnerable to a resource exhaustion issue within the MiniMax M3 VL video processor . The root cause is the lack of range validation or min_pixels
EUVD-2026-125169
vLLM from 0.24.0 before 0.31.0 contains a resource exhaustion vulnerability in the MiniMax M3 VL video processor, which accepts request-level minpixels without range or minpixels = maxpixels validation. Its video-wide pixel budget is enforced only when maxlongsidepixel is set, and that value...
CVE-2026-126396 vLLM from 0.24.0 before 0.31.0 Resource Exhaustion via MiniMax M3 min_pixels
vLLM from 0.24.0 before 0.31.0 contains a resource exhaustion vulnerability in the MiniMax M3 VL video processor, which accepts request-level minpixels without range or minpixels = maxpixels validation. Its video-wide pixel budget is enforced only when maxlongsidepixel is set, and that value...
EUVD-2026-125168
vLLM from 0.10.1 before 0.31.0 forwards a request-controlled mmprocessorkwargs.maxpatches value to the Hugging Face processor for Cohere2 Vision models without a server-side upper bound, causing quadratic CPU work during multimodal preprocessing. Remote clients can send small image chat-completio...
CVE-2026-126398
The vLLM library (versions 0.28.0 to 0.30.0 ) contains a reachable assertion vulnerability in the GPUModelRunner._init_xdrope_positions component. This flaw allows remote attackers to cause a Denial of Service (DoS) by submitting specific embeddings-only requests. Specifically, an attacker can se...
EUVD-2026-125167
vLLM from 0.28.0 before 0.30.0 contains a reachable assertion vulnerability in GPUModelRunner.initxdropepositions that allows remote attackers to crash the engine by submitting embeddings-only requests. Attackers can send a /v1/completions request containing only promptembeds to XD-RoPE models wi...
CVE-2026-126399
The vLLM framework versions 0.19.0 before 0.31.0 is susceptible to a resource exhaustion vulnerability. An inference client can bypass server frame limits by manipulating the media_io_kwargs.video.num_frames parameter in chat completion requests. By providing a large value or -1 alongside multipl...
EUVD-2026-125166
vLLM from 0.19.0 before 0.31.0 contains a resource exhaustion vulnerability that allows inference clients to bypass the server frame limit by setting mediaiokwargs.video.numframes in chat completion requests. Attackers can supply numframes=-1 or a large value with many base64 JPEG frames, forcing...
LeakGauge
LeakGauge This is the code repository for our paper: The Model's Tell: Measuring Context-Leakage Attack Signals with Behavior Gauges. ArXiv version and paper link: https://arxiv.org/abs/2608.17829 This repository implements the LeakGauge pipeline for leakage detection: demo dataset preparation,...
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...
chatml-role-token-probe
chatml-role-token-probe Prove your agent stack is vulnerable to role-forgery — before an attacker does. A drop-in probe that fires a corpus of ChatML / Mistral / Llama control-token literals at your self-hosted agent stack's untrusted-text path and reports whether the tokenizer or server turns th...
ai-kubernetes-helm-chart-security
AI on Kubernetes: Helm chart security defaults evidence This repository holds the data and evidence behind the Sorami technical report "AI on Kubernetes: 10 of 14 Helm charts ship with no API auth" version 1.2. Read the report here: https://sorami.com.au/research/ai-kubernetes-helm-chart-security...
EXPLOIT-CVE-2026-22778
CVE-2026-22778 — vLLM RCE in video processing Vulnerable lab + proof of concept for CVE-2026-22778 CVSS 9.8, an unauthenticated remote code execution chain in vLLM's multimodal ingestion path. | ---|--- CVE| CVE-2026-22778 Advisory| GHSA-4r2x-xpjr-7cvv Affected| vLLM = 0.8.3, vLLM turned...
CVE-2026-22807_Range
🛡️ CVE-2026-22807 Simulation: AI Supply Chain RCE Target Range ⚠️ Disclaimer This project is intended solely for security research and educational purposes , aiming to demonstrate the supply chain attack principle Remote Code Execution during AI model loading. Do not use the generated malicious...