PyPI/vllm

vLLM affected by RCE via auto_map dynamic module loading during model initialization

Modified: 9/10/2026

vLLM denial of service via prompt embeds on M-RoPE models

Modified: 9/10/2026

vLLM is vulnerable to Server-Side Request Forgery (SSRF) through `MediaConnector` class

Modified: 9/10/2026

vLLM: Unauthenticated OOM Denial of Service via Unbounded `n` Parameter in OpenAI API Server

Modified: 9/10/2026

vLLM's Artifact Pin Decay allows pinned deployments to load unpinned code, weights, and processors

Modified: 9/10/2026

vLLM: ReDoS via structured_outputs.regex in the lm-format-enforcer backend (no compile timeout) — missed sibling of GHSA-rwxx-mrjm-wc2m

Modified: 9/10/2026

vLLM: SSRF + arbitrary local file read in MiMoV2OmniMultiModalProcessor `_fetch_image` and audio loader bypass MediaConnector protections

Modified: 9/10/2026

Potential Timing Side-Channel Vulnerability in vLLM’s Chunk-Based Prefix Caching

Modified: 8/7/2026

vLLM has RCE In Video Processing

Modified: 9/10/2026

vLLM: GGUF dequantize kernel int truncation exposes uninitialized GPU memory in multi-tenant serving

Modified: 9/10/2026

vLLM Deserialization of Untrusted Data vulnerability

Modified: 8/7/2026

vLLM vulnerable to DoS via large Chat Completion or Tokenization requests with specially crafted `chat_template_kwargs`

Modified: 9/10/2026

vLLM: Processing differential in multi-channel audio downmixing enables hidden-input/moderation bypass for audio models

Modified: 7/17/2026

vLLM: Resource-Exhaustion (DoS) through Malicious Jinja Template in OpenAI-Compatible Server

Modified: 9/10/2026

vLLM: OOM Denial of Service via Audio Decompression Bomb

Modified: 9/10/2026

vLLM DOS: Remotely kill vllm over http with invalid JSON schema

Modified: 8/7/2026

vLLM has Hardcoded Trust Override in Model Files Enables RCE Despite Explicit User Opt-Out

Modified: 9/10/2026

vLLM has remote code execution vulnerability in the tool call parser for Qwen3-Coder

Modified: 9/10/2026

vLLM: temperature=NaN and temperature=Infinity bypass validation and propagate to GPU kernels

Modified: 9/10/2026

vLLM: Cross-User Data Leak Vulnerability

Modified: 9/10/2026

vLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parameters

Modified: 9/10/2026

vLLM: Derender endpoints decode caller-supplied GenerateResponse token IDs without output bounds

Modified: 9/10/2026

vLLM: Completion prompt lists fan out into unbounded engine requests

Modified: 9/10/2026

vLLM vulnerable to remote code execution via transformers_utils/get_config

Modified: 7/17/2026

vLLM: image EXIF Rotation & PNG tRNS Transparency Not Normalized, Causing Mismatch Between Model Input and Expectations

Modified: 9/10/2026

vLLM: Request-selected PyNvVideoCodec GPU decode bypasses static VRAM reservation

Modified: 9/17/2026

vLLM has Remote DoS via Invalid Recovered Token Reinjection

Modified: 9/10/2026

vLLM: OpenAI auth bypass

Modified: 9/10/2026

vllm has Improper Resource Shutdown or Release

Modified: 7/13/2026

Data exposure via ZeroMQ on multi-node vLLM deployment

Modified: 8/7/2026

vLLM allows clients to crash the openai server with invalid regex

Modified: 8/7/2026

Remote Code Execution Vulnerability in vLLM Multi-Node Cluster Configuration

Modified: 8/7/2026

vLLM has a Weakness in MultiModalHasher Image Hashing Implementation

Modified: 8/7/2026

vLLM allows Remote Code Execution by Pickle Deserialization via AsyncEngineRPCServer() RPC server entrypoints

Modified: 8/7/2026

CVE-2025-24357 Malicious model remote code execution fix bypass with PyTorch < 2.6.0

Modified: 8/7/2026

vLLM is vulnerable to DoS in Idefics3 vision models via image payload with ambiguous dimensions

Modified: 9/10/2026

vLLM: Unauthenticated audio decompression-bomb DoS in /v1/chat/completions

Modified: 9/16/2026

vLLM vulnerable to Denial of Service by abusing xgrammar cache

Modified: 4/15/2025

vLLM: incomplete CVE-2026-22778 fix leaks PIL repr addresses via Anthropic router

Modified: 9/10/2026

vLLM Vulnerable to Remote Code Execution via Mooncake Integration

Modified: 8/7/2026

vLLM Allows Remote Code Execution via PyNcclPipe Communication Service

Modified: 8/7/2026

vLLM Vulnerable to Remote DoS via Special-Token Placeholders

Modified: 9/10/2026

vLLM: Unauthenticated Internal Path and Username Disclosure via Validation Error Messages

Modified: 9/10/2026

vLLM vulnerable to Regular Expression Denial of Service

Modified: 8/7/2026

vLLM introduced enhanced protection for CVE-2025-62164

Modified: 9/11/2026

vLLM denial of service via outlines unbounded cache on disk

Modified: 8/7/2026

vLLM deserialization vulnerability leading to DoS and potential RCE

Modified: 9/10/2026

vLLM: Server-Side Request Forgery (SSRF) in `download_bytes_from_url `

Modified: 9/10/2026

vLLM deserialization vulnerability in vllm.distributed.GroupCoordinator.recv_object

Modified: 8/7/2026

vLLM vulnerable to DoS with incorrect shape of multimodal embedding inputs

Modified: 9/10/2026

vLLM: Denial of Service via Unbounded Frame Count in video/jpeg Base64 Processing

Modified: 9/10/2026

vLLM: Incomplete CVE-2025-62164 remediation can be bypassed by concurrent prompt parts

Modified: 9/10/2026

vLLM: Security Check Bypass via assert Statement in Activation Function Loading Allows Arbitrary Code Execution

Modified: 9/10/2026

vLLM vulnerable to Server-Side Request Forgery (SSRF) through MediaConnector

Modified: 9/10/2026

vllm: Malicious model to RCE by torch.load in hf_model_weights_iterator

Modified: 8/7/2026

vLLM uses Python 3.12 built-in hash() which leads to predictable hash collisions in prefix cache

Modified: 8/7/2026

vLLM: ReDoS via structured_outputs.regex compiled without timeout in xgrammar and outlines backends

Modified: 9/10/2026

vllm API endpoints vulnerable to Denial of Service Attacks

Modified: 9/10/2026

vLLM has SSRF Protection Bypass

Modified: 7/17/2026

vLLM: Speech-to-text upload size limit is enforced after full UploadFile read

Modified: 7/17/2026

vLLM: Quadratic Time Complexity in Input Token Processing​ leads to denial of service

Modified: 8/7/2026

vLLM Tool Schema allows DoS via Malformed pattern and type Fields

Modified: 8/7/2026

vLLM denial of service vulnerability

Modified: 8/7/2026

vLLM has a Regular Expression Denial of Service (ReDoS, Exponential Complexity) Vulnerability in `pythonic_tool_parser.py`

Modified: 8/7/2026

vLLM Denial of Service via the best_of parameter

Modified: 8/7/2026

vLLM is vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method

Modified: 8/18/2026

vLLM is vulnerable to timing attack at bearer auth

Modified: 9/10/2026

vLLM makes Use of Uninitialized Resource

Modified: 9/10/2026

vLLM Allows Remote Code Execution via Mooncake Integration

Modified: 8/7/2026

Modified: 6/10/2026

Modified: 5/20/2026

Modified: 6/10/2026

Modified: 5/29/2025

Modified: 6/10/2026

Modified: 6/26/2025

Modified: 6/26/2025

Modified: 6/26/2025

Modified: 6/27/2025

Modified: 7/1/2025

Modified: 7/1/2025

Modified: 5/20/2026

Modified: 5/20/2026

Modified: 5/20/2026

vLLM affected by RCE via auto_map dynamic module loading during model initialization

Modified: 7/7/2026

vLLM is vulnerable to Server-Side Request Forgery (SSRF) through `MediaConnector` class

Modified: 7/7/2026

vLLM vulnerable to DoS via large Chat Completion or Tokenization requests with specially crafted `chat_template_kwargs`

Modified: 7/7/2026

vLLM: Resource-Exhaustion (DoS) through Malicious Jinja Template in OpenAI-Compatible Server

Modified: 7/7/2026

vLLM has remote code execution vulnerability in the tool call parser for Qwen3-Coder

Modified: 7/7/2026

vLLM vulnerable to remote code execution via transformers_utils/get_config

Modified: 7/7/2026

Data exposure via ZeroMQ on multi-node vLLM deployment

Modified: 7/7/2026

Remote Code Execution Vulnerability in vLLM Multi-Node Cluster Configuration

Modified: 7/7/2026

vLLM deserialization vulnerability leading to DoS and potential RCE

Modified: 7/7/2026

vLLM vulnerable to DoS with incorrect shape of multimodal embedding inputs

Modified: 7/7/2026

vLLM vulnerable to Server-Side Request Forgery (SSRF) through MediaConnector

Modified: 7/7/2026

vllm API endpoints vulnerable to Denial of Service Attacks

Modified: 7/7/2026

phi4mm: Quadratic Time Complexity in Input Token Processing​ leads to denial of service

Modified: 7/7/2026

vLLM Tool Schema allows DoS via Malformed pattern and type Fields

Modified: 7/7/2026

vLLM denial of service vulnerability

Modified: 7/7/2026

vLLM Denial of Service via the best_of parameter

Modified: 7/7/2026

vLLM is vulnerable to timing attack at bearer auth

Modified: 7/7/2026