CVE-2026-73556

Received Received - Intake

Regex Denial of Service in vLLM Structured Output Engine

Publication date: 2026-08-13

Last updated on: 2026-08-14

Assigner: GitHub, Inc.

Description

vLLM is an inference and serving engine for large language models. Prior to 0.26.0, the structured_outputs.regex parameter in vllm/v1/structured_output/backend_lm_format_enforcer.py is passed to lmformatenforcer.RegexParser without compile_regex_with_timeout or validation in validate_structured_output_request_lm_format_enforcer, allowing an unauthenticated /v1/completions request against the lm-format-enforcer backend to consume a CPU core and stall the structured-output engine path with a catastrophic regular expression. This issue is fixed in version 0.26.0.

CVSS Scores

EPSS Scores

Probability:
Percentile:

Meta Information

Published

2026-08-13

Last Modified

2026-08-14

Generated

2026-09-02

AI Q&A

2026-08-13

EPSS Evaluated

2026-09-01

NVD

CVE-2026-73556

EUVD

EUVD-2026-58063

Affected Vendors & Products

Vendor	Product	Version / Range
vllm-project	vllm	0.26.0

Helpful Resources

Exploitability

CWE

KEV

CWE ID	Description
CWE-1333	The product uses a regular expression with an inefficient, possibly exponential worst-case computational complexity that consumes excessive CPU cycles.
CWE-400	The product does not properly control the allocation and maintenance of a limited resource.

Attack-Flow Graph

Executive Summary

This is a Regular Expression Denial of Service (ReDoS) vulnerability in the vLLM project's lm-format-enforcer backend. It allows an unauthenticated request to submit a malicious regex pattern that consumes excessive CPU resources, stalling the structured-output engine and causing a denial of service.

Detection Guidance

To detect this vulnerability, check if your vLLM instance is running a version prior to 0.26.0. Use commands like 'pip show vllm' or 'vllm --version' to verify the installed version. If the version is below 0.26.0, the system is vulnerable.

Impact Analysis

If exploited, this vulnerability can cause system slowdowns or crashes by consuming all available CPU resources with a single malicious request. This reduces service availability and may require manual intervention to restore normal operation.

Compliance Impact

This vulnerability primarily impacts availability by allowing resource exhaustion through ReDoS attacks, which could lead to service disruptions. While not directly violating GDPR or HIPAA, prolonged downtime may indirectly affect compliance by impairing access to personal or health data processing systems. Organizations using affected vLLM versions must ensure timely patching to maintain service availability and data processing integrity.

Mitigation Strategies

Upgrade vLLM to version 0.26.0 or later to apply the regex timeout protection and validation fixes in the lm-format-enforcer backend.

If using the lm-format-enforcer backend, ensure it is configured with compile_regex_with_timeout and early regex pattern validation as implemented in the fix.

Hi! I’m here to help you understand CVE-2026-73556. Ask me anything about the vulnerability, its impact, or mitigation strategies.

0/70