Casdoor CVE-2026-9095

Casdoor must be configured as a SAML 2.0 Service Provider with at least one SAML IdP enabled - deployments using only OAuth, OIDC, LDAP, or local accounts are not exposed via this code path. The attacker must obtain a previously valid SAML <Response>/<Assertion> for the target user (capture vectors: TLS-terminating proxy, log files containing SAMLResponse, browser cache on shared host, malicious co-tenant SP federated to the same IdP, or IdP-side disclosure). No password, MFA token, or user interaction is required at the replay stage, and the assertion remains replayable indefinitely because no ID cache or OneTimeUse check is performed; the only natural limit is the NotOnOrAfter timestamp inside the assertion itself, which is IdP-configured and often hours long.

The CVSS 8.1 score (AV:N/AC:H/PR:N/UI:N/C:H/I:H/A:H) reflects unauthenticated network-based total compromise, tempered by High attack complexity because the adversary must first obtain a valid SAML assertion (e.g., via TLS interception, log exposure, IdP-side leakage, or a malicious relying party in a shared IdP scenario). No EPSS score, KEV listing, or public POC was provided, so exploitation appears opportunistic rather than mass-automated at present. The intelligence package contains a conflict worth flagging: the input includes a 'Denial Of Service' tag, but the description and CVSS vector clearly describe an authentication-bypass/account-takeover impact - the DoS tag appears mislabeled and should be disregarded in favor of the credential-replay impact. Real-world priority is HIGH for any Casdoor deployment using SAML for SSO into privileged systems, because a single captured assertion yields persistent admin access bypassing MFA.

An attacker who captures a single SAML assertion - for example, by harvesting it from server logs, browser history on a shared workstation, an intercepted proxy, or a malicious sibling SP sharing the same IdP - POSTs that same assertion to the Casdoor /api/acs (Assertion Consumer Service) endpoint and receives an authenticated session as the original subject, including administrators. No public exploit code is identified at time of analysis, but reproduction is straightforward using standard SAML tooling such as SAMLRaider or a simple curl replay of the captured SAMLResponse parameter. The High attack complexity rating reflects the assertion-capture step, not the replay itself, which is trivial.