CVE-2026-66788 in Advanced Cluster Management for Kubernetes
Summary
by MITRE • 08/20/2026
A flaw was found in Lighthouse. A remote attacker, by compromising a spoke cluster, can exploit a vulnerability where the destination namespace for resource injection is derived from an attacker-controlled label or annotation on the broker object. This allows the attacker to inject unauthorized EndpointSlices and ServiceImports into any namespace on peer clusters, including critical system namespaces like kube-system and openshift-*. This could lead to privilege escalation or other forms of system compromise within the cluster.
If you want to get the best quality for vulnerability data then you always have to consider VulDB.
Analysis
by VulDB Data Team • 08/20/2026
The vulnerability identified in Lighthouse represents a critical security flaw rooted in improper validation of input data during resource injection processes. Specifically, the destination namespace for injected resources is dynamically derived from labels or annotations present on broker objects rather than being strictly controlled by administrative policy or internal state verification. This design choice creates an opportunity for remote attackers who have achieved initial access to a spoke cluster within a multi-cluster management topology. By compromising such a node or service account with sufficient permissions, the attacker can manipulate these metadata fields to dictate where subsequent resources are deployed across the federated environment. The core technical flaw lies in the lack of strict allow-listing or namespace isolation checks when processing broker object attributes, allowing external input to directly influence critical infrastructure configuration paths.
The operational impact of this vulnerability is severe due to its potential for lateral movement and privilege escalation within Kubernetes-based environments. An attacker can exploit this mechanism to inject unauthorized EndpointSlices and ServiceImports into any target namespace on peer clusters connected via the Lighthouse federation layer. This capability effectively bypasses standard network segmentation and access control boundaries that rely on namespace isolation. By injecting these specific resource types, an attacker can manipulate service discovery mechanisms, potentially redirecting traffic to malicious endpoints or disrupting legitimate services. The ability to write into critical system namespaces such as kube-system and openshift-* is particularly dangerous, as it grants the potential for complete cluster compromise, including the manipulation of core DNS resolution, load balancing configurations, and internal service routing that underpins application availability and security posture.
From a classification perspective, this vulnerability aligns with CWE-94 Improper Control of Generation of Code (Code Injection) in its broader sense of injecting unauthorized objects into system-managed spaces, but more accurately maps to CWE-732 Incorrect Permission Assignment for Critical Resource due to the failure to restrict write access based on proper authorization checks. In terms of adversary behavior, this exploitation technique corresponds to ATT&CK tactic T1098 Unauthorized Account Modification and specifically sub-technique T1098.004 SSH Authorized Keys Manipulation if applied to identity contexts, or more generally T1526 Cloud Service Discovery followed by T1530 Data from Cloud Storage Object for reconnaissance and persistence. The attack vector leverages the trust relationship between clusters in a federation setup, turning a feature designed for seamless service discovery into a conduit for unauthorized administrative actions across cluster boundaries.
Mitigation strategies must focus on enforcing strict validation of broker object metadata before it is used to determine resource placement. Administrators should implement rigorous input sanitization that rejects any labels or annotations not explicitly defined in an approved schema. Additionally, role-based access control policies within the Lighthouse broker and peer clusters should be tightened to prevent non-administrative service accounts from modifying critical federation resources. Network-level controls such as network policies can help limit the blast radius by restricting communication between namespaces, although this does not address the root cause of unauthorized injection. Regular auditing of broker object configurations and monitoring for anomalous changes in EndpointSlice or ServiceImport objects across peer clusters are essential defensive measures to detect exploitation attempts early.