Multicluster Engine EUVD-2026-57439

Exploitation requires an authenticated principal (PR:L) that already possesses rights to create a ClusterCurator custom resource within a namespace on an MCE hub cluster running the cluster-curator-controller. The concrete trigger named in the description is creating a ClusterCurator resource with a specific naming convention that causes the controller to create a cluster-scoped ClusterRoleBinding. Limiting factors: the attacker must have a valid cluster identity with namespace-level create permission on ClusterCurator objects (this is not remote-unauthenticated), the cluster-curator-controller must be deployed and reconciling (i.e., MCE/ACM with cluster curation in use), and exploitation is most meaningful in multi-tenant clusters where namespace isolation is the security boundary. No user interaction, no non-standard client, and no social engineering are needed once that namespace foothold exists.

The CVSS 3.1 vector CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H (9.9 Critical) is internally consistent with the description: a low-privileged authenticated user (PR:L) acting over the Kubernetes API (AV:N) with low complexity and no user interaction triggers a scope change (S:C) yielding full confidentiality, integrity and availability impact. This is a genuine priority in any multi-tenant MCE deployment because the barrier - the ability to create one custom resource in a namespace you already control - is trivial for an insider or for an attacker who has compromised any tenant workload with namespace API access. No EPSS score, KEV listing, or POC was provided in the input, so exploitation-probability signals are absent; the absence of KEV means there is no confirmed active exploitation, and the absence of a POC means no public exploit is identified at time of analysis. The main mitigating factor is that this is not a remote unauthenticated internet-facing bug: it requires an existing authenticated foothold with rights to create a ClusterCurator, so real-world risk concentrates in shared/multi-tenant clusters rather than single-tenant installations.

A developer or CI service account holding only namespace-scoped access to an MCE hub cluster creates a ClusterCurator resource named according to the specific convention the controller trusts. The cluster-curator-controller reconciles it and creates a cluster-scoped ClusterRoleBinding granting the attacker cluster-wide rights, after which they read platform secrets and delete hosted clusters or node pools. No user interaction or elevated starting privileges are required and no public POC is identified at time of analysis.