Cloud Foundry UAA CVE-2026-40965

The target must be a Cloud Foundry UAA instance running uaa_release v76.12.0-v78.12.0 (or cf-deployment v30.0.0-v56.0.0) AND configured to use EC (Elliptic Curve) keys for JWT token signing - RSA-signed deployments are not affected, which is a meaningful limiter because RSA is the historical default. Given that precondition, no other special conditions apply: the /token_keys endpoint is unauthenticated and network-reachable by design (CVSS AV:N/AC:L/PR:N/UI:N), so any attacker who can reach the UAA HTTPS port can retrieve the key with one request. No user interaction, no admin role, and no special client are required.

Signals are consistent and severe: CVSS 3.1 vector AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:L describes a network-reachable, no-auth, no-interaction flaw with scope change, which fits exactly - the leaked key compromises every resource server that trusts UAA-signed JWTs, not just UAA itself. No EPSS, KEV, or SSVC data was supplied, and there is no public exploit identified at time of analysis, so we cannot confirm in-the-wild abuse. However, the trigger is a single unauthenticated GET against a well-known endpoint, so the gap between disclosure and mass-scanning is expected to be small; this is a real priority for any operator running an affected version with EC signing keys. The one limiting factor is configuration: deployments still using the default RSA signing keys are not exposed, which materially narrows the affected population compared to the raw 10.0 score.

An unauthenticated attacker on the network discovers an affected UAA instance, issues a single HTTP GET to https://<uaa-host>/token_keys, and parses the returned JWK Set to extract the EC private key's 'd' parameter. Using that private key, the attacker mints a JWT impersonating any user or client (including platform admins) with arbitrary scopes and presents it to any service in the foundation that trusts UAA-signed tokens, achieving full platform compromise. No public exploit identified at time of analysis, but the steps are trivial enough to script in minutes once a vulnerable target is identified.