Advisories for Pypi/Vllm package

Current vLLM main lets an inference request choose the PyNvVideoCodec GPU video decoder through media_io_kwargs.video.video_backend, but engine GPU memory reservation is computed only from static startup configuration and VLLM_VIDEO_LOADER_BACKEND. If the server starts with the default OpenCV/software backend and no –mm-ipc-gpu-memory-gb budget, a client can still route a video request into the PyNvVideoCodec path after startup, causing frontend CUDA-context, decoder-surface, and decoded-frame GPU allocations that were not carved out of …

The audio decode-duration guard (max_duration_s, env VLLM_MAX_AUDIO_DECODE_DURATION_S, default 600s) that protects against audio decompression-bomb DoS is wired into only the speech-to-text path (/v1/audio/transcriptions). The chat audio path (/v1/chat/completions, input_audio content parts) calls the same decoder with no limit, so an unauthenticated client can submit a few-KB compressed audio file that expands to multiple GB of float32 PCM at decode time, OOM-killing the worker. This is a distinct sibling of CVE-2026-5497 …

vllm/transformers_utils/processors/mimo_v2_omni.py — the multimodal processor for MiMoV2OmniForCausalLM — issues requests.get(…) directly on user-supplied image and audio URL strings and Image.open(…) on user-supplied local paths, without the SSRF / allowed_local_media_path checks that vllm.multimodal.utils.MediaConnector was hardened with in GHSA-qh4c-xf7m-gxfc, GHSA-v359-jj2v-j536, and GHSA-pf3h-qjgv-vcpr. This is the same bug class as those three published advisories, in a code path the patches missed. When a user passes a URL or local-file string through multi_modal_data (e.g. …

An integer overflow in the act_and_mul_kernel kernel can cause the output of one user request to be incorporated into the response of another request within the same inference batch. Under certain conditions, the last request in a batch can receive a partial or complete copy of the first user's inference result, resulting in cross-user data leakage.

When the vLLM API receives a malformed request (e.g., invalid JSON or missing required fields), FastAPI raises a Pydantic RequestValidationError. The validation_exception_handler in vllm/entrypoints/openai/server_utils.py converts this exception to a string via str(exc), which includes the internal file path and line number of the handler function. The existing sanitize_message() function in vllm/entrypoints/utils.py strips memory addresses (e.g., 0x7f…) but does not strip File "…", line X patterns. The result is a user-facing …

The fix for GHSA-rwxx-mrjm-wc2m ("ReDoS via structured_outputs.regex compiled without timeout") wrapped the regex compile in the xgrammar and outlines backends with compile_regex_with_timeout (and, for outlines, validate_regex_is_buildable). The lm-format-enforcer backend was left unguarded: it compiles the attacker-supplied regex with no timeout and no buildability check. A single request with a catastrophic regex hangs the structured-output compile step and stalls the engine worker (denial of service).

The /v1/completions/derender and /v1/chat/completions/derender endpoints accept caller-supplied GenerateResponse objects and postprocess every nested choices[*].token_ids list directly. Unlike the normal render/generate path, derender does not enforce model context length, resolved max_tokens, max_num_seqs, choice-count, or response-size bounds before detokenizing and returning the supplied token IDs. An authenticated API client can therefore make the CPU-only render frontend, or any server exposing these /v1 derender routes, spend CPU and memory proportional to attacker-chosen generated-output-shaped …

The follow-up protection for CVE-2025-62164 is incomplete at vLLM revision 26587f9519e22a5c4549ead7595ad9ca3229c4fd. It wraps serialized prompt-embedding reconstruction and dense conversion in torch.sparse.check_sparse_tensor_invariants(), but PyTorch 2.11.0 implements that context with save/enable/restore operations over process-global state. Two prompt-embedding parts in one /v1/chat/completions request are gathered concurrently on the event loop's default executor. When one context exits before the other loads its tensor, it can restore the global flag to False while the second …

The /v1/completions request model accepts prompt as a list of text prompts or a list of token-id prompts without any outer prompt-count bound. The serving path turns each element into a separate engine input, creates one engine generator per element, merges all generators, and allocates a response slot per prompt. An authenticated API client can therefore turn one request into an attacker-chosen number of backend subrequests before any aggregate request-count …

Short summary of the problem. Make the impact and severity as clear as possible. For example: An unsafe deserialization vulnerability allows any unauthenticated user to execute arbitrary code on the server. Sending a pure prompt embeds payload in a /v1/completions request with a model using M-RoPE causes the EngineCore to fail an assertion and fatally crash, shutting down the entire server application. Any remote user who is authorized to make …

Current-head vLLM documents VLLM_MAX_AUDIO_CLIP_FILESIZE_MB as the maximum audio file size accepted by the speech-to-text APIs. The default is 25 MB. vllm/envs.py also describes files larger than this value as rejected. The /v1/audio/transcriptions and /v1/audio/translations routes call await request.file.read() before vLLM checks that limit. In FastAPI and Starlette, UploadFile.read() returns bytes from the uploaded file object; when called without a size argument, the route materializes the remaining file contents. vLLM then …

The structured_outputs.regex API parameter passes a user-supplied regex string directly to grammar compiler backends with no compilation timeout. In the xgrammar backend, the string reaches compile_regex() with no guard. In the outlines backend, validate_regex_is_buildable() blocks structural issues (lookarounds, backreferences) but provides zero protection against exponential DFA state-space explosion. Patterns like (a+)+b pass all checks and hang the inference worker.

Issue Description Librosa defaults to using numpy.mean for mono downmixing (to_mono), while the international standard ITU-R BS.775-4 specifies a weighted downmixing algorithm. This discrepancy results in: Inconsistency between audio heard by humans (e.g., through headphones/regular speakers) and audio processed by AI models (Which infra via Librosa, such as vllm, transformer). https://github.com/librosa/librosa/blob/af8c839fb15317fa2712ea66e7a22da6a9267b32/librosa/core/audio.py#L478 Attack Scenario and Impact LFE (Low-Frequency Effects) Channel Exploit Attackers can craft special multichannel audio files containing: Normal content …

A frontend-legal multi-request speculative workload can make vLLM produce an out-of-vocabulary recovered token equal to vocab_size, convert that value to -1 when choosing the next live token for a request, and then feed that -1 back into the next drafter input ids. On Qwen3 GPTQ this reaches the worker-side drafting / attention path and crashes the engine with a GPU device-side assert. The same issue is reachable through the public …

All temperature validation gates use comparison operators (<, >), which silently evaluate to False for NaN and for positive Infinity in Python's IEEE 754 float semantics. Both values pass every guard and propagate to GPU sampling kernels, where they produce undefined behavior or CUDA errors that can crash the inference worker. Note: -Infinity is correctly caught.

vLLM's /v1/audio/transcriptions endpoint limits compressed upload size but not decoded PCM output. A 25MB OPUS file expands to ~14.9GB of float32 PCM at decode time. Tested on vLLM v0.19.0.

The fix for CVE-2026-22778 / GHSA-4r2x-xpjr-7cvv (PRs #31987 and #32319) introduced sanitize_message and applied it at four FastAPI exception-handling sites in the OpenAI router. The sanitizer strips object-repr memory addresses (<_io.BytesIO object at 0x7a95e299e750> → <_io.BytesIO object>) before error messages reach the client, defeating the ASLR-bypass primitive that CVE-2026-22778 chained with a libopenjp2 heap overflow for RCE. The fix is incomplete: response paths added to vLLM at or after the …

Issue 1: EXIF orientation not normalized → The image orientation processed by the model differs from how humans view it, introducing interpretation bias. Issue 2: PNG tRNS not explicitly flattened before converting to RGB → After conversion, transparent/semi-transparent pixels are rendered unexpectedly, making otherwise subtle overlay elements visible and distorting the input content. (This attack is similar to AlphaDog: RGBA handling is already correct in vLLM, but since tRNS permits …

Integer truncation of tensor dimensions in vLLM's GGUF dequantize kernels (csrc/quantization/gguf/gguf_kernel.cu) causes partial tensor processing. The output tensor is allocated at full size via torch::empty (uninitialized memory), but the dequantize CUDA kernel processes only a truncated number of elements. The unfilled portion of the output tensor retains whatever was previously in GPU memory. In multi-tenant inference deployments, this residual GPU memory may contain tensor data from other users' inference requests, …

An assert-based security check in vLLM's activation function loading allows any unauthenticated attacker to achieve arbitrary code execution on the server by publishing a malicious HuggingFace model, when vLLM runs in Python optimized mode (python -O or PYTHONOPTIMIZE=1).

A vulnerability in ASGI web servers and starlette's trust on those web servers enables an authentication bypass of the OpenAI API AuthenticationMiddleware, which was discovered during @x41sec's source code audit. It allows to use the API without providing the configured VLLM_API_KEY or –api-key.

vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the VideoMediaIO.load_base64() method. When processing video/jpeg data URLs, the method splits the base64 data string on commas to extract individual JPEG frames without enforcing a frame count limit. An attacker can exploit this by crafting a single API request containing thousands of comma-separated base64-encoded JPEG frames in …

vLLM's revision pinning controls do not consistently apply to all artifacts loaded for a model. A deployment that supplies –revision or –code-revision can still load dynamic code, GGUF files, image processors, retrieval side weights, or same-repository subfolder weights/config from an unpinned/default revision. This is a supply-chain integrity issue for pinned vLLM deployments. Operators can believe they are serving a reviewed model revision while vLLM resolves behavior-affecting nested or sibling artifacts …

A vulnerability was identified in vllm-project vllm 0.19.0. This issue affects some unknown processing of the component OpenAI-compatible Serving Path. Such manipulation leads to denial of service. It is possible to launch the attack remotely. The exploit is publicly available and might be used. The pull request to fix this issue awaits acceptance.

The extract_hidden_states speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The crash is triggered when any request in the batch uses sampling penalty parameters (repetition_penalty, frequency_penalty, or presence_penalty). A single request with a penalty parameter (e.g., "repetition_penalty": 1.1) is sufficient to crash the server. The crash is deterministic and immediate — no concurrency, …

This report explains a Token Injection vulnerability in vLLM’s multimodal processing. Unauthenticated, text-only prompts that spell special tokens are interpreted as control. Image and video placeholder sequences supplied without matching data cause vLLM to index into empty grids during input-position computation, raising an unhandled IndexError and terminating the worker or degrading availability. Multimodal paths that rely on image_grid_thw/video_grid_thw are affected. Severity: High (remote DoS). Reproduced on vLLM 0.10.0 with Qwen2.5-VL.

A vulnerability was found in vLLM up to 0.19.0. The affected element is the function has_mamba_layers of the file vllm/v1/kv_cache_interface.py of the component KV Block Handler. Performing a manipulation results in uninitialized resource. It is possible to initiate the attack remotely. The attack is considered to have high complexity. The exploitability is described as difficult. The exploit has been made public and could be used. The patch is named 1ad67864c0c20f167929e64c875f5c28e1aad9fd. …

A Denial of Service vulnerability exists in the vLLM OpenAI-compatible API server. Due to the lack of an upper bound validation on the n parameter in the ChatCompletionRequest and CompletionRequest Pydantic models, an unauthenticated attacker can send a single HTTP request with an astronomically large n value. This completely blocks the Python asyncio event loop and causes immediate Out-Of-Memory crashes by allocating millions of request object copies in the heap …

A Server Side Request Forgery (SSRF) vulnerability in download_bytes_from_url allows any actor who can control batch input JSON to make the vLLM batch runner issue arbitrary HTTP/HTTPS requests from the server, without any URL validation or domain restrictions. This can be used to target internal services (e.g. cloud metadata endpoints or internal HTTP APIs) reachable from the vLLM host.

The VideoMediaIO.load_base64() method at vllm/multimodal/media/video.py:51-62 splits video/jpeg data URLs by comma to extract individual JPEG frames, but does not enforce a frame count limit. The num_frames parameter (default: 32), which is enforced by the load_bytes() code path at line 47-48, is completely bypassed in the video/jpeg base64 path. An attacker can send a single API request containing thousands of comma-separated base64-encoded JPEG frames, causing the server to decode all frames …

Two model implementation files hardcode trust_remote_code=True when loading sub-components, bypassing the user's explicit –trust-remote-code=False security opt-out. This enables remote code execution via malicious model repositories even when the user has explicitly disabled remote code trust.

The SSRF protection fix for https://github.com/vllm-project/vllm/security/advisories/GHSA-qh4c-xf7m-gxfc can be bypassed in the load_from_url_async method due to inconsistent URL parsing behavior between the validation layer and the actual HTTP client.

A chain of vulnerabilities in vLLM allow Remote Code Execution (RCE): Info Leak - PIL error messages expose memory addresses, bypassing ASLR Heap Overflow - JPEG2000 decoder in OpenCV/FFmpeg has a heap overflow that lets us hijack code execution Result: Send a malicious video URL to vLLM Completions or Invocations for a video model -> Execute arbitrary commands on the server Completely default vLLM instance directly from pip, or docker, …

A Server-Side Request Forgery (SSRF) vulnerability exists in the MediaConnector class within the vLLM project's multimodal feature set. The load_from_url and load_from_url_async methods obtain and process media from URLs provided by users, using different Python parsing libraries when restricting the target host. These two parsing libraries have different interpretations of backslashes, which allows the host name restriction to be bypassed. This allows an attacker to coerce the vLLM server into …

vLLM loads Hugging Face auto_map dynamic modules during model resolution without gating on trust_remote_code, allowing attacker-controlled Python code in a model repo/path to execute at server startup.

Users can crash the vLLM engine serving multimodal models that use the Idefics3 vision model implementation by sending a specially crafted 1x1 pixel image. This causes a tensor dimension mismatch that results in an unhandled runtime error, leading to complete server termination.

The fix here for CVE-2025-62164 is not sufficient. The fix only disables prompt embeds by default rather than addressing the root cause, so the DoS vulnerability remains when the feature is enabled.

vllm has a critical remote code execution vector in a config class named Nemotron_Nano_VL_Config. When vllm loads a model config that contains an auto_map entry, the config class resolves that mapping with get_class_from_dynamic_module(…) and immediately instantiates the returned class. This fetches and executes Python from the remote repository referenced in the auto_map string. Crucially, this happens even when the caller explicitly sets trust_remote_code=False in vllm.transformers_utils.config.get_config. In practice, an attacker can …

Users can crash the vLLM engine serving multimodal models by passing multimodal embedding inputs with correct ndim but incorrect shape (e.g. hidden dimension is wrong), regardless of whether the model is intended to support such inputs (as defined in the Supported Models page). The issue has existed ever since we added support for image embedding inputs, i.e. #6613 (released in v0.5.5)

The /v1/chat/completions and /tokenize endpoints allow a chat_template_kwargs request parameter that is used in the code before it is properly validated against the chat template. With the right chat_template_kwargs parameters, it is possible to block processing of the API server for long periods of time, delaying all other requests

A memory corruption vulnerability that leading to a crash (denial-of-service) and potentially remote code execution (RCE) exists in vLLM versions 0.10.2 and later, in the Completions API endpoint. When processing user-supplied prompt embeddings, the endpoint loads serialized tensors using torch.load() without sufficient validation. Due to a change introduced in PyTorch 2.8.0, sparse tensor integrity checks are disabled by default. As a result, maliciously crafted tensors can bypass internal bounds checks …

A resource-exhaustion (denial-of-service) vulnerability exists in multiple endpoints of the OpenAI-Compatible Server due to the ability to specify Jinja templates via the chat_template and chat_template_kwargs parameters. If an attacker can supply these parameters to the API, they can cause a service outage by exhausting CPU and/or memory resources.

The API key support in vLLM performed validation using a method that was vulnerable to a timing attack. This could potentially allow an attacker to discover a valid API key using an approach more efficient than brute force.

A Server-Side Request Forgery (SSRF) vulnerability exists in the MediaConnector class within the vLLM project's multimodal feature set. The load_from_url and load_from_url_async methods fetch and process media from user-provided URLs without adequate restrictions on the target hosts. This allows an attacker to coerce the vLLM server into making arbitrary requests to internal network resources. This vulnerability is particularly critical in containerized environments like llm-d, where a compromised vLLM pod could …

An unsafe deserialization vulnerability allows any authenticated user to execute arbitrary code on the server if they are able to get the model to pass the code as an argument to a tool call.

A Denial of Service (DoS) vulnerability can be triggered by sending a single HTTP GET request with an extremely large header to an HTTP endpoint. This results in server memory exhaustion, potentially leading to a crash or unresponsiveness. The attack does not require authentication, making it exploitable by any remote user.

A recent review identified several regular expressions in the vllm codebase that are susceptible to Regular Expression Denial of Service (ReDoS) attacks. These patterns, if fed with crafted or malicious input, may cause severe performance degradation due to catastrophic backtracking.

The vLLM backend used with the /v1/chat/completions OpenAPI endpoint fails to validate unexpected or malformed input in the "pattern" and "type" fields when the tools functionality is invoked. These inputs are not validated before being compiled or parsed, causing a crash of the inference worker with a single request. The worker will remain down until it is restarted.

In the file vllm/multimodal/hasher.py, the MultiModalHasher class has a security and data integrity issue in its image hashing method. Currently, it serializes PIL.Image.Image objects using only obj.tobytes(), which returns only the raw pixel data, without including metadata such as the image’s shape (width, height, mode). As a result, two images of different sizes (e.g., 30x100 and 100x30) with the same pixel byte sequence could generate the same hash value. This …

A Regular Expression Denial of Service (ReDoS) vulnerability exists in the file vllm/entrypoints/openai/tool_parsers/pythonic_tool_parser.py of the vLLM project. The root cause is the use of a highly complex and nested regular expression for tool call detection, which can be exploited by an attacker to cause severe performance degradation or make the service unavailable.

Hitting the /v1/completions API with a invalid json_schema as a Guided Param will kill the vllm server

A denial of service bug caused the vLLM server to crash if an invalid regex was provided while using structured output. This vulnerability is similar to GHSA-6qc9-v4r8-22xg, but for regex instead of a JSON schema. Issue with more details: https://github.com/vllm-project/vllm/issues/17313

This issue arises from the prefix caching mechanism, which may expose the system to a timing side-channel attack.

This advisory duplicates another.

This advisory duplicates another.

This advisory duplicates another.

This advisory duplicates another.

vLLM supports the use of the PyNcclPipe class to establish a peer-to-peer communication domain for data transmission between distributed nodes. The GPU-side KV-Cache transmission is implemented through the PyNcclCommunicator class, while CPU-side control message passing is handled via the send_obj and recv_obj methods on the CPU side.​ A remote code execution vulnerability exists in the PyNcclPipe service. Attackers can exploit this by sending malicious serialized data to gain server control …

In a multi-node vLLM deployment using the V0 engine, vLLM uses ZeroMQ for some multi-node communication purposes. The secondary vLLM hosts open a SUB ZeroMQ socket and connect to an XPUB socket on the primary vLLM host.

A critical performance vulnerability has been identified in the input preprocessing logic of the multimodal tokenizer. The code dynamically replaces placeholder tokens (e.g., <|audio_|>, <|image_|>) with repeated tokens based on precomputed lengths. Due to ​​inefficient list concatenation operations​​, the algorithm exhibits ​​quadratic time complexity (O(n²))​​, allowing malicious actors to trigger resource exhaustion via specially crafted inputs.

vLLM integration with mooncake is vaulnerable to remote code execution due to using pickle based serialization over unsecured ZeroMQ sockets. The vulnerable sockets were set to listen on all network interfaces, increasing the likelihood that an attacker is able to reach the vulnerable ZeroMQ sockets to carry out an attack. This is a similar to GHSA - x3m8 - f7g5 - qhm7, the problem is in

This advisory duplicates another.

In a multi-node vLLM deployment, vLLM uses ZeroMQ for some multi-node communication purposes. The primary vLLM host opens an XPUB ZeroMQ socket and binds it to ALL interfaces. While the socket is always opened for a multi-node deployment, it is only used when doing tensor parallelism across multiple hosts. Any client with network access to this host can connect to this XPUB socket unless its port is blocked by a …

https://github.com/vllm-project/vllm/security/advisories/GHSA-rh4j-5rhw-hr54 reported a vulnerability where loading a malicious model could result in code execution on the vllm host. The fix applied to specify weights_only=True to calls to torch.load() did not solve the problem prior to PyTorch 2.6.0. PyTorch has issued a new CVE about this problem: https://github.com/advisories/GHSA-53q9-r3pm-6pq6 This means that versions of vLLM using PyTorch before 2.6.0 are vulnerable to this problem.

This report is to highlight a vulnerability in XGrammar, a library used by the structured output feature in vLLM. The XGrammar advisory is here: https://github.com/mlc-ai/xgrammar/security/advisories/GHSA-389x-67px-mjg3 The xgrammar library is the default backend used by vLLM to support structured output (a.k.a. guided decoding). Xgrammar provides a required, built-in cache for its compiled grammars stored in RAM. xgrammar is available by default through the OpenAI compatible API server with both the V0 …

vllm-project vllm version 0.6.0 contains a vulnerability in the distributed training API. The function vllm.distributed.GroupCoordinator.recv_object() deserializes received object bytes using pickle.loads() without sanitization, leading to a remote code execution vulnerability.

vllm-project vllm version v0.6.2 contains a vulnerability in the MessageQueue.dequeue() API function. The function uses pickle.loads to parse received sockets directly, leading to a remote code execution vulnerability. An attacker can exploit this by sending a malicious payload to the MessageQueue, causing the victim's machine to execute arbitrary code.

vllm-project vllm version 0.6.0 contains a vulnerability in the AsyncEngineRPCServer() RPC server entrypoints. The core functionality run_server_loop() calls the function _make_handler_coro(), which directly uses cloudpickle.loads() on received messages without any sanitization. This can result in remote code execution by deserializing malicious pickle data.

The outlines library is one of the backends used by vLLM to support structured output (a.k.a. guided decoding). Outlines provides an optional cache for its compiled grammars on the local filesystem. This cache has been on by default in vLLM. Outlines is also available by default through the OpenAI compatible API server.

When vLLM is configured to use Mooncake, unsafe deserialization exposed directly over ZMQ/TCP will allow attackers to execute remote code on distributed hosts.

Maliciously constructed prompts can lead to hash collisions, resulting in prefix cache reuse, which can interfere with subsequent responses and cause unintended behavior.

This advisory duplicates another.

The vllm/model_executor/weight_utils.py implements hf_model_weights_iterator to load the model checkpoint, which is downloaded from huggingface. It use torch.load function and weights_only parameter is default value False. There is a security warning on https://pytorch.org/docs/stable/generated/torch.load.html, when torch.load load a malicious pickle data it will execute arbitrary code during unpickling.

A flaw was found in the vLLM library. A completions API request with an empty prompt will crash the vLLM API server, resulting in a denial of service.

A vulnerability was found in the ilab model serve component, where improper handling of the best_of parameter in the vllm JSON web API can lead to a Denial of Service (DoS). The API used for LLM-based sentence or chat completion accepts a best_of parameter to return the best completion from several options. When this parameter is set to a large value, the API does not handle timeouts or resource exhaustion …