CVE-2026-9799: Keycloak UMA Authorization Bypass in Permissive Mode

Technical summary

CVE-2026-9799 is an authorization bypass in org.keycloak.authorization affecting Red Hat Build of Keycloak. An authenticated principal possessing a UMA permission ticket for resource A can craft a permission request using a specific prefix that bypasses per-resource access control checks, granting access to resources B, C, etc. of the same type within the resource server. The flaw arises from insufficient validation of permission request parameters when resources are typed, owner-managed, and lack explicit protective policies. Exploitation requires: (1) authentication; (2) existing UMA grant for at least one resource; (3) PERMISSIVE policy enforcement mode enabled; and (4) target resources configured with ownerManagedAccess without hardened policy guards. CWE-639 (Authorization Bypass Through User-Controlled Key) maps to the root cause.

Business impact

The vulnerability permits authenticated insiders or compromised accounts to access or modify resources beyond their intended scope. In multi-tenant or data-segregated environments using UMA, this could lead to cross-resource data leakage or unauthorized edits. The medium CVSS score reflects the requirement for authentication and user interaction, as well as the confidentiality and integrity impact. Availability is not affected. Organizations using Keycloak for fine-grained access control—particularly those managing sensitive documents, healthcare records, or financial data across resource types—face elevated risk if their deployment matches the vulnerable configuration profile.

Affected systems

Red Hat Build of Keycloak is explicitly affected. The vulnerability is configuration-dependent: it manifests only when (1) the resource server uses PERMISSIVE policy enforcement mode, (2) resources are typed with ownerManagedAccess enabled, and (3) no explicit authorization policy protects the resource type. Installations using ENFORCING mode, untyped resources, or resources with hardened policies are not vulnerable. Organizations should audit their Keycloak resource server configurations and policy enforcement settings to determine exposure.

Exploitability

Exploitation requires authentication and valid UMA credentials for at least one resource. An attacker must know or discover the permission request prefix pattern that triggers the bypass and must target a resource server in the vulnerable configuration state. The CVSS vector (AC:L, PR:L, UI:R) indicates low attack complexity, low privilege requirements (authenticated user suffices), and user interaction (likely a legitimate permission request that the user initiates). Practical exploitation is moderately straightforward for an insider with UMA access but requires specific knowledge of the prefix mechanism and target resource server configuration.

Remediation

Apply a security patch from Red Hat that remediates the authorization validation logic in org.keycloak.authorization. Red Hat typically releases patches through the standard update channels. Additionally, as a compensating control, configure affected resource servers to use ENFORCING policy enforcement mode instead of PERMISSIVE, and ensure all typed resources with ownerManagedAccess are protected by explicit authorization policies. Organizations should verify their UMA configurations and test access control boundaries in lower environments before deployment.

Patch guidance

Verify availability of a patched Red Hat Build of Keycloak release that addresses this authorization bypass. Consult the Red Hat security advisory corresponding to CVE-2026-9799 for specific version numbers and update instructions. Test patches in a staging environment to confirm UMA functionality and policy behavior remain intact. Prioritize patching for production Keycloak instances configured with PERMISSIVE mode on typed, owner-managed resources.

Detection guidance

Monitor Keycloak audit logs for permission requests with unusual prefixes or requests that succeed across resource boundaries when issued by a single principal. Audit the effective resource server policies to identify which are in PERMISSIVE mode with typed, unprotected resources. Review UMA permission ticket grants and cross-reference with actual resource access patterns for anomalies. Implement alerting on failed permission checks that subsequently succeed, which may indicate exploitation attempts. Validate that authorization policies are properly evaluated by performing permission request tests from controlled test accounts.

Why prioritize this

Although the CVSS score is medium (4.6), prioritization depends on deployment profile. Organizations running Keycloak with PERMISSIVE enforcement on sensitive typed resources should treat this as high-priority; those using ENFORCING mode or without ownerManagedAccess are lower-priority. The flaw permits lateral privilege escalation within a resource server and data exfiltration, making it material for any organization relying on Keycloak for access control enforcement. Early patching is recommended given the authentication-free nature of post-exploitation access.

Risk score, explained

CVSS 3.1 base score 4.6 (MEDIUM) reflects: (1) network-accessible attack vector (AV:N); (2) low attack complexity requiring only crafted permission requests (AC:L); (3) low privilege threshold—any authenticated user suffices (PR:L); (4) user interaction requirement, suggesting a user-initiated permission workflow (UI:R); (5) confidentiality impact from unauthorized resource reading (C:L); (6) integrity impact from unauthorized resource modification (I:L); (7) no availability impact (A:N); (8) unchanged scope (S:U). The score is moderate rather than high due to authentication and configuration prerequisites, but organizations in vulnerable configurations should elevate their internal risk assessment accordingly.

Frequently asked questions