Summary
Kyverno automatically includes the admission controller's ServiceAccount token in outbound HTTP requests made through apiCall service mode when no explicit authorization header is provided. If an attacker can influence the destination of such a request, the token may be exfiltrated and used to control Kyverno policies or access cluster resources.
Affected context
The vulnerability can occur in Kyverno deployments that use the apiCall service mode for outbound HTTP requests and rely on the admission controller's default request behavior. Risk is increased when request destinations or policy inputs can be influenced by untrusted users, tenants, or external data, allowing sensitive authentication material to reach an external or attacker-controlled endpoint.
Exploit preconditions
- Attack surface: Kyverno policy rules that issue outbound apiCall requests
- Required access: Ability to influence a policy, request parameter, or other input that controls the outbound destination
- Trigger point: Processing an apiCall request without an explicit authorization header
- Why exploitation is possible: The admission controller automatically attaches its ServiceAccount token to the outbound request, potentially exposing a credential to the selected destination
Why this happens
The underlying issue is an unsafe authentication-header default in an outbound request feature. When a caller does not explicitly provide an authorization header, Kyverno automatically forwards the admission controller's ServiceAccount token instead of treating the request as unauthenticated or requiring an explicit opt-in. This behavior maps to CWE-200: Exposure of Sensitive Information to an Unauthorized Actor, because a credential is disclosed across a trust boundary to a destination that may not be authorized to receive it.
Potential impact
- An attacker may obtain the admission controller's ServiceAccount token
- The exposed token could be used to impersonate the Kyverno controller
- An attacker could modify or disable Kyverno policies
- An attacker could access or alter cluster resources permitted by the ServiceAccount
- Policy enforcement, admission decisions, and workload integrity could be affected
General mitigation guidance
Mitigation generally focuses on preventing implicit credential forwarding and reducing trust in outbound destinations.
Common defensive patterns include:
- Treating outbound requests as unauthenticated unless credential use is explicitly configured and authorized
- Preventing automatic propagation of ServiceAccount tokens to external services
- Restricting apiCall destinations through allowlists and network egress controls
- Limiting which users and policies can configure outbound request targets
- Applying least-privilege permissions to the Kyverno ServiceAccount
- Using dedicated, narrowly scoped credentials when an external service requires authentication
- Avoiding sensitive data transmission to destinations that are not explicitly trusted
How teams detect issues like this
Organizations can look for this class of issue through:
- Static analysis and policy checks for implicit authentication on outbound HTTP clients
- Code review signals involving default authorization headers or automatic token propagation
- Inventory reviews of Kyverno policies that use apiCall service mode
- Admission and audit log analysis for unexpected external request destinations
- Network monitoring for ServiceAccount token patterns or unusual egress from Kyverno components
- Policy enforcement that restricts outbound destinations and requires explicit approval for credential-bearing requests
Automated security review tools may help identify credential forwarding paths and unrestricted egress configurations.

