A flaw in the User-Managed Access token endpoint of Keycloak allows an attacker to craft a JSON Web Token with a malicious azp claim that is used to set the Access‑Control‑Allow‑Origin header before the token is verified. The header is reflected in the response even if the grant is later rejected, potentially revealing low‑sensitivity error information. The underlying weakness is insufficient validation of user input before header generation.

The vulnerability affects Red Hat Build of Keycloak. No specific version range is listed in the available data, so all installations that have not yet applied a vendor fix are potentially exposed.

With a CVSS score of 3.7, the vulnerability presents a moderate severity risk. The EPSS score is not available and the issue is not listed in CISA’s KEV catalog, indicating no known widespread exploitation to date. The likely attack vector is remote, requiring the attacker to send a specially crafted JWT to the UMA endpoint of a misconfigured client that permits all web origins. Successful exploitation can compromise origin isolation and disclose error responses, but does not grant arbitrary code execution.