Summary
A server-side request forgery vulnerability in Fulcio allows cross-host redirects during OIDC discovery to reach attacker-controlled or unintended endpoints. In affected deployments, Kubernetes ServiceAccount tokens may be attached to these requests, enabling token disclosure, malicious JWKS substitution, or blind SSRF when an issuer is malicious or compromised.
Affected context
The issue occurs in Fulcio deployments that perform OIDC discovery and follow redirects across host boundaries. It becomes risky when the discovery process can access Kubernetes-mounted ServiceAccount tokens, automatically inject those tokens into requests, or load token material from local paths without sufficiently restricting the destination and source.
Exploit preconditions
- Attack surface: Fulcio's OIDC discovery and JWKS retrieval process
- Required access: The ability to control, compromise, or influence an OIDC issuer or its discovery responses
- Trigger point: Redirect handling or token-assisted requests made during OIDC discovery
- Why exploitation is possible: Cross-host redirects and insufficiently constrained token handling may cause Fulcio to contact attacker-controlled hosts or disclose Kubernetes ServiceAccount credentials
Why this happens
The underlying flaw is insufficient destination and credential-boundary enforcement in an OIDC discovery client. The client follows redirects to different hosts and may attach Kubernetes ServiceAccount tokens or load local token material without adequately verifying that the destination is trusted and in scope. This behavior maps to CWE-918: Server-Side Request Forgery, because attacker-influenced URLs cause the server to make requests to unintended locations and may expose privileged request context.
Potential impact
- An attacker may induce requests to internal or external systems through Fulcio
- Kubernetes ServiceAccount tokens could be disclosed to external hosts
- Malicious JWKS keys could be substituted or cached
- OIDC trust decisions and certificate issuance could be influenced
- Compromised tokens could enable further access within the cluster
General mitigation guidance
Mitigation generally focuses on enforcing strict network, redirect, and credential boundaries in OIDC clients.
Common defensive patterns include:
- Rejecting cross-host redirects by default
- Allowing redirects only to explicitly trusted origins
- Injecting ServiceAccount tokens only for approved destinations and request contexts
- Restricting local token loading to narrowly defined, protected paths
- Separating discovery requests from privileged runtime credentials where possible
- Applying least-privilege permissions and network egress controls to Fulcio workloads
- Validating and safely caching JWKS responses with issuer and origin binding
How teams detect issues like this
Organizations can detect related risks through:
- Static analysis and code review checks for HTTP clients that follow redirects without host validation
- Searches for automatic credential or token injection in generic outbound request code
- Review of local file reads used to obtain Kubernetes credentials
- Network monitoring for Fulcio requests to unexpected hosts or external destinations
- Policies requiring explicit allowlists for OIDC discovery and JWKS endpoints
- Alerts for ServiceAccount tokens appearing in outbound request headers or logs
Automated security review tools may help identify redirect handling and sensitive-header propagation patterns.

