CVE-2026-105114

OpenAM before 16.1.3 contains a server-side request forgery vulnerability. Attackers able to register or modify OAuth 2.0 clients can make OpenAM fetch internal resources via an unvalidated jwks_uri. They can trigger unauthenticated fetches to probe internal hosts, metadata endpoints or local files, or exhaust request threads for DoS.

OpenAM before 16.1.3 contains an improper authorization vulnerability that allows delegated administrators to destroy sessions outside their realms because realm checks use the requester's realm. Authenticated accounts holding the iplanet-am-session-destroy-sessions attribute can supply a target session identifier or handle to forcibly log out users in any realm.

OpenAM before 16.1.3 contains an authorization bypass vulnerability in the sessions REST endpoint query operation that allows realm administrators to list sessions of every realm. Attackers holding delegated RealmAdmin privileges can supply a _queryFilter naming another realm to disclose usernames, universal IDs, and session handles across tenant boundaries.

OpenAM before 16.1.3 applies its OAuth2 Provider PKCE enforcement only to authorization requests whose response_type is exactly 'code'. Codes issued through OpenID Connect hybrid flows (code token, code id_token, code token id_token) carry no bound challenge. An attacker who intercepts such a code can redeem it for a public client's tokens with any non-empty code_verifier.

OpenAM before 16.1.3 contains an open redirect vulnerability that allows unauthenticated attackers to redirect users by supplying an unverified id_token_hint to the /oauth2/connect/endSession endpoint. Attackers can name any realm client in a forged hint to redirect victims to any registered post-logout URI, enabling phishing that borrows the OpenAM host's trust.

OpenAM before 16.1.3 contains an email content injection vulnerability that allows unauthenticated attackers to control notification email wording via the forgotPassword and register actions on /json/{realm}/users. Attackers can supply subject and message fields to send phishing mail from the organisation's configured From address, or abuse register as a relay to arbitrary recipients.

OpenAM before 16.1.3 contains a latent cross-site scripting defect that places the SAML message, relay state and target URL unencoded into the load-balancer cookie bounce auto-submit page. If reachable with cookieHashRedirectEnabled set, crafted requests could execute script in the OpenAM origin, though an unrelated HTTP 500 failure prevents exploitation in released versions.

OpenAM before 16.1.3 contains an unauthenticated arbitrary class instantiation vulnerability in the legacy JAX-RPC SOAP interface. Remote attackers can send SOAP requests to /jaxrpc/* with an unverified session identifier and a chosen class name, crashing the server, probing the classpath, or potentially reaching code execution.

OpenAM 10.1.0 has an XXE (XML External Entity) vulnerability in the path /SSOPOST/metaAlias/%realm%/idpv2, allowing remote attackers to read arbitrary files via the SAMLRequest parameter.

In OpenAM prior to 16.1.2, the pre-authentication /authservice PLL endpoint accepts a CustomCallback XML element whose className value selects an arbitrary Java class for AuthXMLUtils to load and instantiate without verifying it implements DSAMECallbackInterface. Default configurations expose the endpoint without authentication, allowing attacker-controlled class initialization and unsafe deserialization to execute code in the server process. Enabling sunRemoteAuthSecurityEnabled does not prevent the vulnerable parsing because its check occurs later.