Summary
OpenChoreo exposes caller-facing management APIs through an externally reachable agent listener without authentication. A network-reachable attacker may invoke proxy and exec operations, access data-plane Kubernetes APIs, and execute commands in workload pods when the affected multi-cluster deployment is reachable.
Affected context
The issue occurs in OpenChoreo's internal cluster gateway when management endpoints are served through the externally reachable agent listener without an effective authentication requirement. It is particularly risky in multi-cluster deployments where network access to the listener permits requests to proxy Kubernetes API traffic or reach workload pod execution functionality.
Exploit preconditions
- Attack surface: Externally reachable agent listener exposing
/api/proxy/and/api/exec/ - Required access: Network connectivity to the affected listener; valid application authentication is not required by the vulnerable path
- Trigger point: Direct requests to caller-facing management API operations
- Why exploitation is possible: The listener routes sensitive management functionality without first enforcing authentication or equivalent authorization
Why this happens
The underlying bug is an authorization boundary failure in which a listener intended for agent communication also exposes management operations without verifying the caller's identity.
This maps primarily to CWE-306: Missing Authentication for Critical Function, because sensitive proxy and command-execution functions can be reached without authentication. It also relates to CWE-668: Exposure of Resource to Wrong Sphere and CWE-862: Missing Authorization, because externally reachable traffic can access operations intended for trusted internal callers.
Potential impact
- Attackers may proxy requests to data-plane Kubernetes APIs
- Attackers could execute commands in workload pods
- Unauthorized users may read, modify, or disrupt cluster-managed resources
- Compromised workloads may enable lateral movement or data access
- Multi-cluster environments can experience broader control-plane or workload impact
General mitigation guidance
Mitigation generally focuses on separating externally reachable transport from sensitive management functionality and enforcing authentication and authorization before dispatch.
Common defensive patterns include:
- Require strong service authentication for every proxy and execution operation
- Apply explicit authorization checks based on endpoint, tenant, cluster, and workload scope
- Bind management APIs to private or mutually authenticated channels where possible
- Use separate listeners or routing domains for agent communication and caller-facing APIs
- Apply least-privilege Kubernetes identities and narrowly scoped pod-execution permissions
- Deny access by default and enforce network policies around gateway and cluster components
- Avoid exposing diagnostic or command-execution endpoints unless they are explicitly required
How teams detect issues like this
Organizations can look for:
- Static analysis and code review signals where HTTP handlers dispatch privileged operations before authentication middleware
- Routes containing proxy, exec, management, or cluster-control functionality without explicit identity and authorization checks
- Listener configurations that expose internal management endpoints on public or broadly reachable interfaces
- Access logs showing unauthenticated requests to
/api/proxy/or/api/exec/ - Kubernetes audit events involving unexpected API proxying, pod command execution, or service-account activity
- Policy checks requiring authentication middleware and least-privilege authorization for sensitive routes

