The vulnerability resides in Dapr Sentry’s OpenID Connect discovery endpoint, which constructs the issuer and jwks_uri fields of the /.well-known/openid-configuration document directly from the request Host header. When an attacker supplies an X-Forwarded-Host header and the service is running with the default configuration—no allowed-hosts list or jwt-issuer specified—the header is accepted without validation. This flaw allows a remote, unauthenticated attacker to poison the discovery document with attacker‑controlled values. Because the document is cached publicly for one hour, client applications that perform dynamic, unpinned discovery will trust the attacker’s JWKS and accept JWTs signed by the attacker, effectively bypassing authentication.

The flaw affects deployments of the Dapr framework, specifically the Dapr Sentry component, in version 1.17.0 and 1.18.0. Any instance running with the default OIDC settings—no jwt-issuer or oidc-allowed-hosts configured— is vulnerable.

This issue carries a CVSS score of 8.2, reflecting high severity. The EPSS score is less than 1%, suggesting a very low probability of exploitation in reality, and it is not listed in the CISA KEV catalog. Attackers can exploit the flaw remotely and without authentication by crafting a malicious X‑Forwarded‑Host header. After poisoning the discovery document, client applications that perform dynamic discovery will treat the attacker’s keys as valid and accept forged tokens, leading to a full authentication bypass.