PYSEC-2026-3865
Vulnerability from pysec - Published: 2026-09-10 09:44 - Updated: 2026-09-10 11:02MLflow's AI Gateway accepts an auth_config.api_base value when creating a gateway secret (mlflow/server/handlers.py, _create_gateway_secret) with no validation of scheme, host, or IP range; the value is stored verbatim. The gateway proxy endpoint (mlflow/server/gateway_api.py, raw_proxy) subsequently issues an HTTP request to that stored api_base plus a caller-supplied path and returns the full response body. MLflow's existing SSRF guard, _validate_webhook_url (which blocks non-global and metadata IPs), is never invoked anywhere in this gateway secret/proxy code path. The CreateGatewaySecret action additionally has no entry in the permission-validator map, so it requires only basic authentication rather than any specific scope, meaning any authenticated user — including read-only accounts — can create a secret pointing at an internal address and reach it via the proxy endpoint, potentially exposing cloud-instance IAM credentials via metadata services. This is related to CVE-2026-4035, which addresses a distinct mechanism in the same gateway-secret feature (server-side $ENV_VAR resolution inside the api_key field leaking credentials to the configured upstream); the finding here is an independent missing-validation gap in the api_base destination itself, unaffected by that fix.
| Name | purl | mlflow | pkg:pypi/mlflow |
|---|
{
"affected": [
{
"package": {
"ecosystem": "PyPI",
"name": "mlflow",
"purl": "pkg:pypi/mlflow"
},
"ranges": [
{
"events": [
{
"introduced": "3.13.0"
},
{
"last_affected": "3.15.2"
}
],
"type": "ECOSYSTEM"
}
],
"versions": [
"3.13.0",
"3.14.0",
"3.15.0",
"3.15.1",
"3.15.2"
]
}
],
"aliases": [
"CVE-2026-71211",
"GHSA-h7x2-h6g9-p789"
],
"details": "MLflow\u0027s AI Gateway accepts an auth_config.api_base value when creating a gateway secret (mlflow/server/handlers.py, _create_gateway_secret) with no validation of scheme, host, or IP range; the value is stored verbatim. The gateway proxy endpoint (mlflow/server/gateway_api.py, raw_proxy) subsequently issues an HTTP request to that stored api_base plus a caller-supplied path and returns the full response body. MLflow\u0027s existing SSRF guard, _validate_webhook_url (which blocks non-global and metadata IPs), is never invoked anywhere in this gateway secret/proxy code path. The CreateGatewaySecret action additionally has no entry in the permission-validator map, so it requires only basic authentication rather than any specific scope, meaning any authenticated user \u2014 including read-only accounts \u2014 can create a secret pointing at an internal address and reach it via the proxy endpoint, potentially exposing cloud-instance IAM credentials via metadata services. This is related to CVE-2026-4035, which addresses a distinct mechanism in the same gateway-secret feature (server-side $ENV_VAR resolution inside the api_key field leaking credentials to the configured upstream); the finding here is an independent missing-validation gap in the api_base destination itself, unaffected by that fix.",
"id": "PYSEC-2026-3865",
"modified": "2026-09-10T11:02:14.447300Z",
"published": "2026-09-10T09:44:51.189114Z",
"references": [
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-71211"
},
{
"type": "WEB",
"url": "https://github.com/mlflow/mlflow/commit/96b7d900e14227e0821981409d86f5fb6c59b386"
},
{
"type": "PACKAGE",
"url": "https://github.com/mlflow/mlflow"
},
{
"type": "WEB",
"url": "https://github.com/mlflow/mlflow/tree/v3.14.0/mlflow"
},
{
"type": "PACKAGE",
"url": "https://pypi.org/project/mlflow"
},
{
"type": "ADVISORY",
"url": "https://github.com/advisories/GHSA-h7x2-h6g9-p789"
}
],
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:N",
"type": "CVSS_V3"
}
],
"summary": " MLflow AI Gateway permits SSRF through an unvalidated api_base"
}
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.