CVE-2026-115917

vLLM: Permissive default CORS origins allow cross-origin requests to the vLLM HTTP server

Published Oct 26, 2026 CVSS 3.1 · 3.4 Low EPSS < 1% KEV · Not listed
Description
vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the OpenAI-compatible server configures Starlette CORSMiddleware with allowed_origins set to a wildcard by default. A malicious web page can use a victim's browser to reach an unauthenticated vLLM server on localhost, an internal network, or a VPN, call endpoints including GET /v1/models and POST /v1/chat/completions, consume GPU resources, and read responses. Cookie-based cross-site request forgery does not apply because credentials are disabled by default, and deployments requiring bearer authentication are exposed only when the caller already has a usable token. This issue is fixed in version 0.32.0.

Analysis

AI-generated

Technical impact

vLLM’s OpenAI‑compatible server originally configured Starlette CORS middleware with a wildcard origin policy. This permissive setting allowed any browser to send requests from arbitrary web pages to the vLLM HTTP endpoints. An attacker could host a malicious webpage, trick a user into visiting it, and then use the victim’s browser to call unauthenticated endpoints such as GET /v1/models and POST /v1/chat/completions. The responses could be read to gain insight into available models and to consume GPU resources. Because credentials are disabled by default, cookie‑based cross‑site request forgery mechanisms do not apply, but the open CORS policy itself is a weakness resembling CWE‑346 and CWE‑942. The impact is primarily data exposure and unauthorized use of compute resources, though no direct remote code execution or privilege escalation is possible.

Affected Systems

Deployments of the vLLM inference engine from vllm‑project with versions prior to 0.32.0 are vulnerable. The issue is fixed in v0.32.0 and subsequent releases where the allowed_origins default has been tightened.

Risk and Exploitability

The risk profile is moderate. The CVSS score of 3.4 indicates low severity, and the EPSS score is unavailable, so exploitation likelihood is uncertain. The vulnerability is not listed in CISA’s KEV catalog. Likely attack vectors involve a compromised or malicious web page accessed by a user who has local or internal network access to the vLLM server. If the server is exposed to the public internet or a VPN, the same page can send requests from any client using the browser, making the exploit practicality higher when users are expected to interact with the service from their browsers.

Remediation

No solution or workaround provided in the CVE record.

OpenCVE Recommended Actions

  • Upgrade vLLM to version 0.32.0 or later, where the default CORS policy has been restricted.
  • Configure the server’s CORS middleware to allow only trusted origins instead of a wildcard.
  • Enforce bearer authentication on all endpoints and reject unauthenticated requests.

Generated by OpenCVE AI on October 27, 2026 at 13:26 UTC.

Tracking

Sign in to view the affected projects.

Advisories

Source ID Title
Github GHSA Github GHSA GHSA-vvpg-rrww-wpmh vLLM: Permissive default CORS origins allow cross-origin requests to the vLLM HTTP server
History
Type Values Removed Values Added
Metrics ssvc

{'options': {'Automatable': 'no', 'Exploitation': 'none', 'Technical Impact': 'partial'}, 'version': '2.0.3'}

Type Values Removed Values Added
CPEs cpe:2.3:a:vllm:vllm:*:*:*:*:*:*:*:*
Type Values Removed Values Added
First Time appeared Vllm
Vllm vllm
Vllm-project
Vllm-project vllm
Vendors & Products Vllm
Vllm vllm
Vllm-project
Vllm-project vllm
Type Values Removed Values Added
Description vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the OpenAI-compatible server configures Starlette CORSMiddleware with allowed_origins set to a wildcard by default. A malicious web page can use a victim's browser to reach an unauthenticated vLLM server on localhost, an internal network, or a VPN, call endpoints including GET /v1/models and POST /v1/chat/completions, consume GPU resources, and read responses. Cookie-based cross-site request forgery does not apply because credentials are disabled by default, and deployments requiring bearer authentication are exposed only when the caller already has a usable token. This issue is fixed in version 0.32.0.
Title vLLM: Permissive default CORS origins allow cross-origin requests to the vLLM HTTP server
Weaknesses CWE-942
References
Metrics cvssV3_1

{'score': 3.4, 'vector': 'CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:C/C:L/I:N/A:N'}

Weaknesses
  • CWE-346

    Origin Validation Error

  • CWE-942

    Permissive Cross-domain Security Policy with Untrusted Domains

Sources

Status
PUBLISHED
Assigner
GitHub_M
Published
Updated
Oct 31, 2026, 1:38 AM
Reserved
Oct 23, 2026, 8:21 PM
Status
Analyzed
Published
Oct 26, 2026, 6:17 PM
Modified
Oct 31, 2026, 2:17 AM
Severity
Low
Public date
Oct 26, 2026, 5:20 PM