WSO2 Identity Server CVE-2025-13475

Exploitation requires a multi-tenant WSO2 deployment - the vendor explicitly states there is no impact if the deployment does not support multi-tenancy. The attacker must be able to register or control a SaaS application in one tenant bearing the same name as a target application in a different tenant (PR:L), and a legitimate user in the victim tenant must actually grant consent to their application (UI:R) for the misapplied consent to be usable. Limiting factors: single-tenant installs are wholly unaffected; the attack hinges on name collision between applications in separate tenants and on victim consent occurring; and it applies only to the affected version ranges (WSO2 IS 5.10.0 <5.10.0.382, WSO2 API-M 3.2.0 <3.2.0.457 and 3.2.1 <3.2.1.76).

Signals are consistent and point to a genuine but conditional, moderate-priority issue rather than an urgent internet-wide threat. The CVSS 3.1 base score is 7.3 (AV:N/AC:L/PR:L/UI:R/S:U/C:H/I:H/A:N): network reachable, low complexity, but requiring some privilege (PR:L, an application/user presence in a tenant) and victim user interaction (UI:R, the user must actually grant consent). EPSS is very low (0.15%, 4th percentile) and CISA SSVC records Exploitation: none, Automatable: no, Technical Impact: partial - all agreeing this is not readily weaponizable at scale. There is no KEV listing and no known POC. The dominant limiting factor is explicit in the description: the vulnerability 'has no impact if the deployment does not support multi-tenancy,' so single-tenant deployments are unaffected. The CVSS uses S:U, but given that the flaw lets one tenant's consent affect applications in a different tenant (a distinct security authority), a scope change (S:C) is arguably more accurate - verify tenant-boundary semantics against WSO2's advisory.

An attacker who can register applications in a tenant on a shared, multi-tenant WSO2 deployment creates a SaaS application using the exact name of a legitimate application in a victim's tenant. When a user in the victim tenant grants OAuth/OpenID consent to their legitimate same-named application, that consent is incorrectly honored for the attacker's application, which can then access and modify the user's data across the tenant boundary. No public POC is identified, and success depends on the low-complexity but interaction-dependent path reflected in the CVSS vector (PR:L/UI:R).