Authlib 1.7.2 and earlier caches OpenID Connect discovery documents without validating the issuer claim or ensuring that the endpoint values originate from the same domain as the discovery endpoint URL. This flaw allows an attacker to serve a malicious discovery document that replaces every endpoint URL—including authorization_endpoint and token_endpoint—with attacker‑controlled URLs. The library accepts and caches these values, effectively redirecting all subsequent authentication traffic to attacker‑controlled servers. The result is potential credential theft or session hijacking when the host application interacts with the redirected endpoints.

The affected product is the Authlib library released by the Authlib project. Versions 1.7.2 and earlier are vulnerable; no other vendors or products are listed as impacted.

The flaw can be exploited whenever an application performs OIDC discovery and receives a poisoned document, which is then cached until invalidated or the library is upgraded. The CVSS score of 9.8 indicates a critical level of severity, while the EPSS score of less than 1% suggests a low probability of exploitation based on current data. The vulnerability is not listed in CISA’s KEV catalog, implying no confirmed exploitation reports. Attackers could obtain a poisoned discovery response by controlling a service that responds to the discovery request; this inferred attack vector is consistent with the described behavior.