CVE-2026-10059: Multicluster Engine ClusterCurator Token Escalation (CVSS 9.1)

Executive Summary

A critical privilege escalation vulnerability (CVE-2026-10059, CVSS 9.1) has been disclosed in the ClusterCurator controller of Red Hat Multicluster Engine for Kubernetes. A tenant administrator with namespace-scoped privileges can exploit this flaw by creating a namespaced ClusterCurator resource, which inadvertently grants the ability to mint cluster-scoped tokens — effectively breaking namespace isolation and achieving cross-cluster privilege escalation.

Vulnerability Details

The ClusterCurator controller in Red Hat Multicluster Engine (MCE) is responsible for managing cluster lifecycle automation — running Ansible jobs, managing cluster provisioning sequences, and coordinating curator-based workflows. The controller processes ClusterCurator custom resources that live in tenant namespaces.

The vulnerability exists because the ClusterCurator controller uses a highly privileged service account to process these namespaced resources, but it does not sufficiently validate whether the requesting tenant should be permitted to trigger cluster-scoped operations. A tenant administrator — who has administrative rights only within their namespace — can create a maliciously crafted ClusterCurator resource that causes the controller to mint a cluster-scoped service account token, granting the tenant admin privileges far beyond their intended namespace boundary.

Attack Chain

1. Attacker has tenant namespace-admin privileges in a Multicluster Engine hub
   (legitimate tenant, compromised account, or low-priv insider)
 
2. Attacker creates a crafted ClusterCurator resource in their namespace
   that references or triggers cluster-scoped token creation
 
3. ClusterCurator controller processes the resource under its privileged
   service account without sufficient boundary enforcement
 
4. The controller mints a cluster-scoped token accessible to the attacker
 
5. Attacker uses the minted token to authenticate as a highly privileged
   identity, gaining access across namespaces and managed clusters

Impact

This flaw is particularly severe in multi-tenant Kubernetes environments where the trust boundary between tenants is expected to be enforced at the namespace level. MCE manages multiple clusters simultaneously, meaning a successful exploit on the hub cluster can cascade to all managed clusters.

Affected Products

Check your MCE version:

# Check MCE operator version
oc get csv -n multicluster-engine | grep multicluster-engine
 
# Check ClusterCurator controller
oc get deployment cluster-curator-controller -n multicluster-engine -o jsonpath='{.spec.template.spec.containers[0].image}'

Remediation

Step 1: Apply Red Hat Security Patch

Apply the available patch from Red Hat immediately:

# Update MCE operator via OLM
oc patch subscription multicluster-engine \
  -n multicluster-engine \
  --type merge \
  -p '{"spec":{"channel":"stable-2.x"}}'
 
# Verify the updated CSV is active
oc get csv -n multicluster-engine | grep multicluster-engine

Consult the Red Hat Security Advisory for the exact fixed version applicable to your MCE release.

Step 2: Audit ClusterCurator Resources

Review existing ClusterCurator resources for unexpected or malicious configurations:

# List all ClusterCurator resources across namespaces
oc get clustercurators -A -o wide
 
# Inspect each curator's spec for unusual job references or configurations
oc get clustercurators -A -o yaml | grep -A 20 "spec:"

Step 3: Review Tenant RBAC

Limit which users can create ClusterCurator resources in tenant namespaces:

# Check who can create ClusterCurators in each namespace
oc auth can-i create clustercurators --as=<user> -n <namespace>
 
# Audit all RoleBindings granting admin in MCE-managed namespaces
oc get rolebindings -A -o wide | grep admin

Step 4: Monitor for Anomalous Token Activity

After patching, review audit logs for signs of prior exploitation:

# In OpenShift, audit logs are at:
# /var/log/kube-apiserver/audit.log (or via the audit log API)
 
# Look for token minting events by the ClusterCurator service account
# Specifically: TokenRequest or ServiceAccountToken creation across unexpected namespaces

Detection Indicators

Mitigation Workaround (Pre-Patch)

If patching cannot be applied immediately:

1. Restrict ClusterCurator creation via RBAC in tenant namespaces — remove the ability for untrusted tenant admins to create ClusterCurator resources
2. Audit existing curators and remove any with unexpected or external job references
3. Monitor the ClusterCurator controller's service account for unusual API activity
4. Apply NetworkPolicy to limit what namespaces the ClusterCurator controller can reach

# Remove ClusterCurator create permission from tenant role if not needed
oc patch role <tenant-role> -n <namespace> --type=json \
  -p='[{"op":"remove","path":"/rules/<index>"}]'
# (Identify the exact index for ClusterCurator rules first)

Context: Red Hat Kubernetes Security Advisories — August 2026

CVE-2026-10059 is one of two critical Kubernetes privilege escalation vulnerabilities disclosed by Red Hat on August 5, 2026. The companion issue, CVE-2026-10090 (CVSS 9.9) in Red Hat ACM's Application Subscription controller, shares a similar attack pattern — namespace-scoped users abusing privileged controllers to gain cluster-level access. Organizations running Red Hat's Kubernetes management stack should assess exposure to both vulnerabilities simultaneously.

References

• NVD — CVE-2026-10059
• Red Hat Security Advisory
• Red Hat Multicluster Engine Documentation

Related Reading

• CVE-2026-10090: Red Hat ACM Helm Channel Privilege Escalation (CVSS 9.9)
• CISA Flags Langflow RCE, Tomcat, and N-central Flaws as Actively Exploited