Description

A flaw was found in the cluster-curator-controller component of multicluster engine (MCE). A tenant with create or update permissions on ClusterCurator resources can inject an arbitrary Job specification. This is possible because the CreateJob() function does not validate user-controlled input when unmarshaling the spec.install.overrideJob raw extension. Successful exploitation allows the injected Job to run with the controller's elevated privileges, leading to arbitrary code execution and privilege escalation, potentially accessing cluster-wide secrets.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "9.9",
  "pubDate": "2026-08-12T20:17:53.650Z",
  "pubdate": "2026-08-12T20:17:53.650Z",
  "executiveSummary": "A privilege escalation vulnerability exists in the cluster-curator-controller component of the multicluster engine (MCE). The vulnerability arises from an improper input validation flaw within the unmarshaling mechanism of the spec.install.overrideJob raw extension. This security deficiency allows a low-privileged tenant possessing create or update permissions on ClusterCurator resources to inject an arbitrary Kubernetes Job specification. Successful exploitation of this flaw enables an attacker to execute arbitrary code with the elevated privileges of the controller service account. The resulting impact includes potential unauthorized access to cluster-wide secrets, compromise of the underlying multicluster infrastructure, and total control over affected cluster resources. Exploitation requires authenticated access with specific resource creation or modification privileges within the targeted environment, coupled with the ability to supply malicious payloads via unvalidated Custom Resource definitions. Because the controller operates with elevated administrative privileges across the managed infrastructure, the risk implications are severe, demanding immediate remediation to prevent malicious state manipulation and unauthorized privilege escalation.",
  "technicalDetails": "The vulnerability resides within the cluster-curator-controller component of the multicluster engine (MCE), specifically inside the execution flow governed by the CreateJob() function. The root cause of the vulnerability is the absence of rigorous validation and sanitization procedures when the application unmarshals user-controlled input derived from the spec.install.overrideJob raw extension within ClusterCurator custom resources. Custom Resource Definitions (CRDs) in Kubernetes often rely on runtime extensions or raw JSON/YAML extensions to allow flexible configurations. However, if the controller parses these raw extensions directly into native object types—such as batch/v1 Job specifications—without validating the schema or enforcing structural constraints against an allowlist, it exposes an attack surface for malformed or malicious object injection.\nThe attack flow proceeds as follows: First, an authenticated tenant actor acquires or utilizes existing create or update permissions targeting ClusterCurator resources within the cluster. Second, the adversary crafts a malicious ClusterCurator resource payload containing a maliciously engineered Job specification embedded within the spec.install.overrideJob raw extension field. Third, the cluster-curator-controller reads the newly created or updated ClusterCurator resource and invokes the vulnerable CreateJob() function. Fourth, the function unmarshals the raw extension directly into a Kubernetes Job object without validating the supplied fields or verifying whether the tenant is authorized to instantiate such configurations. Finally, the Kubernetes API server accepts the resulting Job object because the request originates from the cluster-curator-controller, which executes operations using its own elevated service account credentials.\nPost-exploitation impact is severe. The injected Job executes with the high-privilege permissions assigned to the cluster-curator-controller, bypassing standard multitenant boundaries and role-based access control (RBAC) restrictions enforced on the tenant. This elevated context grants the attacker the ability to execute arbitrary containerized code, pivot across the cluster infrastructure, read sensitive cluster-wide secrets, environment variables, or configuration data, and potentially compromise additional managed clusters connected to the multicluster engine framework."
}