vLLM, an inference and serving engine for large language models (LLMs), has a Regular Expression Denial of Service (ReDoS) vulnerability in the file `vllm/entrypoints/openai/tool_parsers/pythonic_tool_parser.py` of versions 0.6.4 up to but excluding 0.9.0. 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. The pattern contains multiple nested quantifiers, optional groups, and inner repetitions which make it vulnerable to catastrophic backtracking. Version 0.9.0 contains a patch for the issue.

vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Prior to 0.24.0, the structured_outputs.regex API parameter passes a user-supplied regular expression string directly to the grammar compiler backends with no compilation timeout; in the xgrammar backend the string reaches the regex compiler with no guard, and in the outlines backend the validation step blocks structural issues such as lookarounds and backreferences but performs no complexity analysis, so a pattern with nested quantifiers passes all checks and causes exponential state-space expansion, allowing a single request containing an adversarial regex to hang an inference worker indefinitely and deny service. This issue is fixed in version 0.24.0.

vLLM versions >= 0.6.3 and < 0.9.0 contain multiple regular expression denial of service (ReDoS) vulnerabilities. Several regex patterns — in vllm/lora/utils.py, the phi4mini tool parser, and the OpenAI-compatible serving chat endpoint — are susceptible to catastrophic backtracking. An attacker submitting crafted input with nested or repeated structures can trigger severe CPU consumption and performance degradation, resulting in denial of service.

vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. 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. The affected code in vLLM is vllm/model_executor/guided_decoding/outlines_logits_processors.py, which unconditionally uses the cache from outlines. A malicious user can send a stream of very short decoding requests with unique schemas, resulting in an addition to the cache for each request. This can result in a Denial of Service if the filesystem runs out of space. Note that even if vLLM was configured to use a different backend by default, it is still possible to choose outlines on a per-request basis using the guided_decoding_backend key of the extra_body field of the request. This issue applies only to the V0 engine and is fixed in 0.8.0.

vLLM is an inference and serving engine for large language models (LLMs). From 0.1.0 to before 0.10.1.1, 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. This vulnerability is fixed in 0.10.1.1.

vLLM is an inference and serving engine for large language models (LLMs). In version 0.8.0 up to but excluding 0.9.0, 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. Version 0.9.0 fixes the issue.

vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Versions starting from 0.5.2 and prior to 0.8.5 are vulnerable to denial of service and data exposure via ZeroMQ on multi-node vLLM deployment. 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 firewall. Once connected, these arbitrary clients will receive all of the same data broadcasted to all of the secondary vLLM hosts. This data is internal vLLM state information that is not useful to an attacker. By potentially connecting to this socket many times and not reading data published to them, an attacker can also cause a denial of service by slowing down or potentially blocking the publisher. This issue has been patched in version 0.8.5.

vLLM is a library for LLM inference and serving. From 0.12.0 to before 0.24.0, sending a pure prompt embeds payload in a /v1/completions request with a model using M-RoPE causes EngineCore to fail an assertion and fatally crash, shutting down the entire server application. Any remote user who is authorized to make a /v1/completions request can make such a request and induce a crash. This issue is fixed in version 0.24.0.

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 a data URL, causing the server to decode all frames into memory and crash due to excessive memory consumption. This vulnerability is reachable via the OpenAI-compatible chat completions API and does not require authentication.

vLLM is an inference and serving engine for large language models. From 0.22.0 to 0.23.0, the /v1/audio/transcriptions and /v1/audio/translations routes call request.file.read() to fully materialize an uploaded audio file into memory before vLLM checks the documented VLLM_MAX_AUDIO_CLIP_FILESIZE_MB compressed upload size limit (default 25 MB) later in the speech-to-text preprocessing step, so an API caller who can reach those routes can submit an oversized multipart upload and cause vLLM to allocate memory proportional to the uploaded file size before the request is rejected as too large, creating memory pressure or terminating the process depending on deployment resource limits. This issue is fixed in version 0.24.0.

vLLM is an inference and serving engine for large language models (LLMs). Prior to 0.23.1rc0, 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. This vulnerability is fixed in 0.23.1rc0.

vLLM is an inference and serving engine for large language models (LLMs). From 0.18.0 to before 0.20.0, 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. This vulnerability is fixed in 0.20.0.

vLLM is an inference and serving engine for large language models (LLMs). From 0.6.1 to before 0.20.0, there is a 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. This vulnerability is fixed in 0.20.0.

vLLM versions >= 0.10.2 and < 0.13.0 are missing sparse tensor validation in multimodal embeddings processing. Because PyTorch disables sparse tensor invariant checks by default, an attacker can submit crafted embedding requests with malformed (negative or out-of-bounds) tensor indices, when the prompt-embeds feature is enabled, to trigger crashes or resource exhaustion (denial of service), with potential for out-of-bounds/write-what-where memory corruption. This continues CVE-2025-62164, whose prior fix only disabled the feature by default rather than addressing the root cause.

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.

vLLM is an inference and serving engine for large language models (LLMs). From 0.7.0 to before 0.19.0, the VideoMediaIO.load_base64() method at vllm/multimodal/media/video.py 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, 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 into memory and crash with OOM. This vulnerability is fixed in 0.19.0.

vLLM is an inference and serving engine for large language models (LLMs). From 0.1.0 to before 0.19.0, 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 before the request even reaches the scheduling queue. This vulnerability is fixed in 0.19.0.

vLLM is an inference and serving engine for large language models (LLMs). In versions from 0.6.4 to before 0.12.0, 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. This issue has been patched in version 0.12.0.

vLLM is an inference and serving engine for large language models (LLMs). Version 0.8.0 up to but excluding 0.9.0 have a Denial of Service (ReDoS) that causes the vLLM server to crash if an invalid regex was provided while using structured output. This vulnerability is similar to GHSA-6qc9-v4r8-22xg/CVE-2025-48942, but for regex instead of a JSON schema. Version 0.9.0 fixes the issue.

vLLM is an inference and serving engine for large language models (LLMs). From version 0.5.5 to before 0.11.1, 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). This issue has been patched in version 0.11.1.

vLLM is an inference and serving engine for large language models (LLMs). From version 0.5.5 to before 0.11.1, 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. This issue has been patched in version 0.11.1.

vLLM is an inference and serving engine for large language models (LLMs). In versions 0.8.0 up to but excluding 0.9.0, hitting the /v1/completions API with a invalid json_schema as a Guided Param kills the vllm server. This vulnerability is similar GHSA-9hcf-v7m4-6m2j/CVE-2025-48943, but for regex instead of a JSON schema. Version 0.9.0 fixes the issue.

There is a MEDIUM severity vulnerability affecting CPython. Regular expressions that allowed excessive backtracking during tarfile.TarFile header parsing are vulnerable to ReDoS via specifically-crafted tar archives.

Marked is a markdown parser and compiler. Prior to version 4.0.10, the regular expression `block.def` may cause catastrophic backtracking against some strings and lead to a regular expression denial of service (ReDoS). Anyone who runs untrusted markdown through a vulnerable version of marked and does not use a worker with a time limit may be affected. This issue is patched in version 4.0.10. As a workaround, avoid running untrusted markdown through marked or run marked on a worker thread and set a reasonable time limit to prevent draining resources.

A vulnerability was found in the Keycloak-services package. If untrusted data is passed to the SearchQueryUtils method, it could lead to a denial of service (DoS) scenario by exhausting system resources due to a Regex complexity.

An issue has been discovered in GitLab CE/EE affecting all versions starting from 16.9 prior to 16.9.7, starting from 16.10 prior to 16.10.5, and starting from 16.11 prior to 16.11.2. A problem with the processing logic for Discord Integrations Chat Messages can lead to a regular expression DoS attack on the server.

Copyparty is a portable file server. Versions prior to 1.18.9, the filter parameter for the "Recent Uploads" page allows arbitrary RegExes. If this feature is enabled (which is the default), an attacker can craft a filter which deadlocks the server. This is fixed in version 1.18.9.

An issue has been discovered in GitLab affecting all versions starting from 13.9 before 14.8.6, all versions starting from 14.9 before 14.9.4, all versions starting from 14.10 before 14.10.1. GitLab was not correctly handling malicious text in the CI Editor and CI Pipeline details page allowing the attacker to cause uncontrolled resource consumption.

Zohocorp ManageEngine Exchange Reporter Plus through 5721 are vulnerable to ReDOS vulnerability in the search module.

FastAPI Guard is a security library for FastAPI that provides middleware to control IPs, log requests, and detect penetration attempts. fastapi-guard's penetration attempts detection uses regex to scan incoming requests. However, some of the regex patterns used in detection are extremely inefficient and can cause polynomial complexity backtracks when handling specially crafted inputs. This vulnerability is fixed in 3.0.1.

In Django 3.2 before 3.2.20, 4 before 4.1.10, and 4.2 before 4.2.3, EmailValidator and URLValidator are subject to a potential ReDoS (regular expression denial of service) attack via a very large number of domain name labels of emails and URLs.

The urlnorm crate through 0.1.4 for Rust allows Regular Expression Denial of Service (ReDos) via a crafted URL to lib.rs. NOTE: the Supplier disputes this, taking the position that "Slow printing of URLs is not a CVE."

Pattern Redirects in Liferay Portal 7.4.3.48 through 7.4.3.76, and Liferay DXP 7.4 update 48 through 76 allows regular expressions that are vulnerable to ReDoS attacks to be used as patterns, which allows remote attackers to consume an excessive amount of server resources via crafted request URLs.

An issue has been discovered in GitLab CE/EE affecting all versions starting from 10.3 before 15.11.10, all versions starting from 16.0 before 16.0.6, all versions starting from 16.1 before 16.1.1. A Regular Expression Denial of Service was possible via sending crafted payloads to the preview_markdown endpoint.

The git-url-parse crate through 0.4.4 for Rust allows Regular Expression Denial of Service (ReDos) via a crafted URL to normalize_url in lib.rs, a similar issue to CVE-2023-32758 (Python).

An issue has been discovered in GitLab EE affecting all versions before 16.8.6, all versions starting from 16.9 before 16.9.4, all versions starting from 16.10 before 16.10.2. It was possible for an attacker to cause a denial of service using malicious crafted content in a junit test report file.

fast-xml-parser is an open source, pure javascript xml parser. fast-xml-parser allows special characters in entity names, which are not escaped or sanitized. Since the entity name is used for creating a regex for searching and replacing entities in the XML body, an attacker can abuse it for denial of service (DoS) attacks. By crafting an entity name that results in an intentionally bad performing regex and utilizing it in the entity replacement step of the parser, this can cause the parser to stall for an indefinite amount of time. This problem has been resolved in v4.2.4. Users are advised to upgrade. Users unable to upgrade should avoid using DOCTYPE parsing by setting the `processEntities: false` option.

A denial of service vulnerability was identified in GitLab CE/EE, versions 16.7.7 prior to 16.8.6, 16.9 prior to 16.9.4 and 16.10 prior to 16.10.2 which allows an attacker to spike the GitLab instance resources usage resulting in service degradation via chat integration feature.

giturlparse (aka git-url-parse) through 1.2.2, as used in Semgrep 1.5.2 through 1.24.1, is vulnerable to ReDoS (Regular Expression Denial of Service) if parsing untrusted URLs. This might be relevant if Semgrep is analyzing an untrusted package (for example, to check whether it accesses any Git repository at an http:// URL), and that package's author placed a ReDoS attack payload in a URL used by the package.

regex is an implementation of regular expressions for the Rust language. The regex crate features built-in mitigations to prevent denial of service attacks caused by untrusted regexes, or untrusted input matched by trusted regexes. Those (tunable) mitigations already provide sane defaults to prevent attacks. This guarantee is documented and it's considered part of the crate's API. Unfortunately a bug was discovered in the mitigations designed to prevent untrusted regexes to take an arbitrary amount of time during parsing, and it's possible to craft regexes that bypass such mitigations. This makes it possible to perform denial of service attacks by sending specially crafted regexes to services accepting user-controlled, untrusted regexes. All versions of the regex crate before or equal to 1.5.4 are affected by this issue. The fix is include starting from regex 1.5.5. All users accepting user-controlled regexes are recommended to upgrade immediately to the latest version of the regex crate. Unfortunately there is no fixed set of problematic regexes, as there are practically infinite regexes that could be crafted to exploit this vulnerability. Because of this, it us not recommend to deny known problematic regexes.

The Denosaurs emoji package provides emojis for dinosaurs. Starting in version 0.1.0 and prior to version 0.3.0, the reTrimSpace regex has 2nd degree polynomial inefficiency, leading to a delayed response given a big payload. The issue has been patched in 0.3.0. As a workaround, avoid using the `replace`, `unemojify`, or `strip` functions.

An issue has been discovered in GitLab affecting all versions starting from 15.11 before 16.1.5, all versions starting from 16.2 before 16.2.5, all versions starting from 16.3 before 16.3.1. An authenticated user could trigger a denial of service when importing or cloning malicious content.

Koa is expressive middleware for Node.js using ES2017 async functions. Prior to versions 0.21.2, 1.7.1, 2.15.4, and 3.0.0-alpha.3, Koa uses an evil regex to parse the `X-Forwarded-Proto` and `X-Forwarded-Host` HTTP headers. This can be exploited to carry out a Denial-of-Service attack. Versions 0.21.2, 1.7.1, 2.15.4, and 3.0.0-alpha.3 fix the issue.

An issue has been discovered in GitLab affecting all versions starting from 15.11 before 16.1.5, all versions starting from 16.2 before 16.2.5, all versions starting from 16.3 before 16.3.1. An authenticated user could trigger a denial of service when importing or cloning malicious content.

An issue has been discovered in GitLab CE/EE affecting all versions from 12.7 prior to 16.6.6, 16.7 prior to 16.7.4, and 16.8 prior to 16.8.1 It was possible for an attacker to trigger a Regular Expression Denial of Service via a `Cargo.toml` containing maliciously crafted input.

sqlparse is a non-validating SQL parser module for Python. In affected versions the SQL parser contains a regular expression that is vulnerable to ReDoS (Regular Expression Denial of Service). This issue was introduced by commit `e75e358`. The vulnerability may lead to Denial of Service (DoS). This issues has been fixed in sqlparse 0.4.4 by commit `c457abd5f`. Users are advised to upgrade. There are no known workarounds for this issue.

A vulnerability was found in Woorank robots-txt-guard. It has been rated as problematic. Affected by this issue is the function makePathPattern of the file lib/patterns.js. The manipulation of the argument pattern leads to inefficient regular expression complexity. The exploit has been disclosed to the public and may be used. The name of the patch is c03827cd2f9933619c23894ce7c98401ea824020. It is recommended to apply a patch to fix this issue. The identifier of this vulnerability is VDB-217448.

All versions of the package word-wrap are vulnerable to Regular Expression Denial of Service (ReDoS) due to the usage of an insecure regular expression within the result variable.

formula is a math and string formula parser. In versions prior to 3.0.1 crafted user-provided strings to formula's parser might lead to polynomial execution time and a denial of service. Users should upgrade to 3.0.1+. There are no known workarounds for this vulnerability.

Discourse is an open source discussion platform. In affected versions a malicious user can cause a regular expression denial of service using a carefully crafted git URL. This issue is patched in the latest stable, beta and tests-passed versions of Discourse. Users are advised to upgrade. There are no known workarounds for this issue.