litellm-unauth-model-exposure: LiteLLM Proxy - Model Exposure

2026-05-14 LiteLLM Proxy PoC Public

Description

LiteLLM proxy was detected with anonymous access to the model listing API. LiteLLM proxies that ship without `general_settings.master_key` (or with the key disabled) expose the configured upstream model catalog via /v1/models and /model/info — revealing which OpenAI / Anthropic / Azure / Bedrock / local-LLM endpoints are wired in, often along with their litellm_params routing config, and frequently leaving /chat/completions and /embeddings reachable without any bearer token.

PoC

id: litellm-unauth-model-exposure

info:
  name: LiteLLM Proxy - Model Exposure
  author: DevamShah
  severity: medium
  description: |
    LiteLLM proxy was detected with anonymous access to the model listing API. LiteLLM proxies that ship without `general_settings.master_key` (or with the key disabled) expose the configured upstream model catalog via /v1/models and /model/info — revealing which OpenAI / Anthropic / Azure / Bedrock / local-LLM endpoints are wired in, often along with their litellm_params routing config, and frequently leaving /chat/completions and /embeddings reachable without any bearer token.
  impact: |
    Attackers can enumerate the proxy's full model catalog and abuse the upstream accounts for free inference (cost amplification), exfiltrate any prompt logs cached locally, or pivot to internal-only model endpoints exposed via the proxy's routing config.
  remediation: |
    Set `general_settings.master_key: sk-...` in the LiteLLM config, or pass `--api_key` on startup. Enforce per-user virtual keys via the LiteLLM admin API and place the proxy behind authenticated reverse proxy or service mesh.
  reference:
    - https://docs.litellm.ai/docs/proxy/virtual_keys
    - https://docs.litellm.ai/docs/proxy/configs
    - https://github.com/BerriAI/litellm
  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-862
  metadata:
    verified: true
    max-request: 1
    vendor: berriai
    product: litellm
    shodan-query:
      - 'http.title:"LiteLLM API - Swagger UI"'
      - 'http.html:"LiteLLM"'
    fofa-query: title="LiteLLM API"
  tags: litellm,llm,ai,misconfig,exposure

http:
  - method: GET
    path:
      - "{{BaseURL}}/model/info"

    matchers-condition: and
    matchers:
      - type: dsl
        dsl:
          - 'status_code == 200'
          - 'contains(content_type, "application/json")'
          - 'contains_all(body, "\"data\":", "\"litellm_params\":", "\"model_info\":")'
        condition: and

    extractors:
      - type: json
        name: models
        part: body
        json:
          - '.data[].model_name'
# digest: 490a0046304402202bc2e4f46c3679a3593827cfb57af522f7a0ae72ccc48322131876fcfc1495b20220287dc9603bd4cd21986f201edd7a252ac189cfca7484adc5b1cb8779d6d3d0aa:922c64590222798bb761d5b6d8e72950

# Visit https://trap.biu.life/ to view exploit trends for this vulnerability.

References

Related Vulnerabilities