Description

A flaw was found in multicluster-observability-addon. This vulnerability allows a managed-cluster identity to reference configuration resources outside its designated namespace. This can lead to the disclosure of sensitive hub Secrets to an attacker-controlled managed cluster.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "7.7",
  "pubDate": "2026-09-11T05:16:38.557Z",
  "pubdate": "2026-09-11T05:16:38.557Z",
  "executiveSummary": "The multicluster-observability-addon is affected by a critical configuration validation vulnerability involving inadequate namespace isolation. This flaw enables a managed-cluster identity to bypass security boundaries and reference configuration resources residing outside its authorized namespace scope.\nThe vulnerability functions as a cross-namespace access control failure, allowing an attacker controlling a compromised or malicious managed cluster to traverse the hub's Kubernetes resource hierarchy. By requesting resources from restricted namespaces, an attacker can exfiltrate sensitive information, most notably hub-resident Secrets.\nThe impact is significant, as it permits unauthorized access to credentials and configuration data meant for hub-level operations. This bypass essentially breaks the multi-tenancy trust model enforced by the hub-spoke architecture. Successful exploitation allows for privilege escalation, as stolen Secrets can provide further access to the hub cluster, downstream managed clusters, or external cloud provider infrastructure associated with those credentials. No specific user interaction is required from the hub administrator; the vulnerability is triggered through the standard resource synchronization or observation request mechanisms inherent to the addon's functionality. The risk is high given that secrets frequently contain sensitive API tokens, private keys, or credentials necessary for cluster orchestration.",
  "technicalDetails": "The root cause of this vulnerability lies in the insufficient validation of namespace scopes within the multicluster-observability-addon request processing logic. When a managed cluster submits a request to the hub-side observability component, the component fails to strictly enforce namespace confinement, allowing the caller to reference arbitrary resources across the hub's Kubernetes cluster.\nIn a secure configuration, the managed-cluster identity should be constrained by an RBAC policy or an admission controller that limits resource access exclusively to a dedicated namespace corresponding to that specific managed cluster. However, the flaw in the addon allows the identity to craft requests that reference cross-namespace paths or labels.\nThe attack flow follows these distinct steps: First, an attacker with control over a managed cluster identifies the API endpoint provided by the hub for observability-related resource synchronization. Second, the attacker crafts a malicious payload that includes a reference to a configuration resource—such as a ConfigMap or secret—located in a namespace reserved for the hub controller or a different tenant. Third, the vulnerability is triggered when the observability addon's controller processes this request without verifying that the resource reference matches the requester's assigned namespace. Finally, the controller retrieves the requested sensitive data and transmits it back to the managed cluster as part of the observability payload.\nThe component responsible for this failure is the reconciliation or retrieval logic within the multicluster-observability-addon that interacts with the Kubernetes API server. By failing to perform a check to validate that the requested object's metadata.namespace property matches the caller's authorized namespace, the system blindly trusts the incoming reference.\nPost-exploitation, an attacker can systematically scan the hub cluster for valuable secrets by iterating through known namespaces and resource naming conventions. Given the nature of Kubernetes secrets, this can result in the full compromise of the hub's control plane identity, lateral movement into other managed clusters connected to the hub, and the exposure of sensitive telemetry or credential data. The vulnerability is present due to a logic flaw in trust boundary enforcement rather than a memory corruption or injection flaw, meaning it is inherent in the design of the cross-namespace resource access mechanism."
}