Detects a vLLM inference server exposing its OpenAI-compatible API without authentication. The /v1/models endpoint returns the list of served models and is reachable by any unauthenticated client, allowing model enumeration and unauthorized inference (resource abuse / cost and information exposure). vLLM does not enable an API key by default; it is only enforced when the server is started with --api-key.
id: vllm-openai-api-exposed
info:
name: vLLM OpenAI-Compatible API - Unauthenticated Exposure
author: fsoppelsa
severity: medium
description: |
Detects a vLLM inference server exposing its OpenAI-compatible API without authentication. The /v1/models endpoint returns the list of served models and is reachable by any unauthenticated client, allowing model enumeration and unauthorized inference (resource abuse / cost and information exposure). vLLM does not enable an API key by default; it is only enforced when the server is started with --api-key.
reference:
- https://docs.vllm.ai/en/latest/serving/openai_compatible_server.html
- https://github.com/vllm-project/vllm
classification:
cvss-metrics: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:L
cvss-score: 6.5
cwe-id: CWE-306
metadata:
verified: true
max-request: 1
shodan-query: title:"vLLM"
fofa-query: title="vLLM"
tags: exposure,vllm,llm,ai,api,unauth,misconfig
http:
- method: GET
path:
- "{{BaseURL}}/v1/models"
matchers-condition: and
matchers:
- type: regex
regex:
- '"object"\s*:\s*"list"'
- type: regex
regex:
- '"owned_by"\s*:\s*"vllm'
- '"max_model_len"'
condition: or
- type: word
part: header
words:
- "application/json"
- type: status
status:
- 200
extractors:
- type: json
part: body
json:
- ".data[].id"
# digest: 4a0a004730450220567f1c4e7e8dea8284a2c802133f90f6fb381ead2d718f1056a18a42f1e1e451022100d5205e2a3b0d7dfb406f16e2995a43d1186fc1af82258bb16d8ca00aa9b1e032:922c64590222798bb761d5b6d8e72950
# Visit https://trap.biu.life/ to view exploit trends for this vulnerability.