On this page (7 sections)
CVE-2026-122302 is a critical privilege escalation flaw in Kyverno, the Kubernetes policy engine. According to the GitHub advisory and the NVD record, versions before 1.19.2 let a namespace admin create resources beyond what their RBAC permissions allow.
If you run Kyverno on a shared or multi-tenant cluster and delegate namespace admin rights to teams, customers or CI systems, this one deserves prompt attention. The public information is short, so this article separates what is confirmed from what is not.
What happened
The advisory was published on November 5, 2026. It says Kyverno's admission checks for generate rules verify the permissions of the Kyverno controller rather than those of the person who wrote the policy. A namespace admin can therefore get the controller to do things the admin could not do directly.
The GitHub entry is marked unreviewed and lists no package, affected range or patched version. The NVD text only repeats the same description, so it is not independent confirmation. A VulnCheck advisory and a project advisory (GHSA-q62c-r9v2-cf64) are referenced, but their contents were not available to us.
How the weakness works
Kyverno can generate resources: a generate rule tells its background controller to create or clone objects, such as a NetworkPolicy or ConfigMap, when something happens in the cluster. The controller usually holds broad permissions so it can do this.
The flaw is a classic confused deputy problem. When a policy is submitted, the check looks at what the controller is allowed to do instead of what the author is allowed to do. Per the advisory, an attacker with namespace admin rights can create a namespaced Policy whose generate rules target other namespaces or cluster-scoped kinds. The background controller then creates or clones those resources cluster-wide, using its own privileges.
The advisory does not list which resource kinds are reachable, so assume anything the controller's service account can write is in scope until the maintainers say otherwise. We are not providing exploit details.
Who is exposed
The description points to someone who can create a namespaced Policy, which in practice means a namespace admin or someone with equivalent rights. That makes these environments most at risk:
- Multi-tenant clusters where teams or customers administer their own namespaces.
- Platforms that give CI/CD pipelines or service accounts rights to manage Policy objects.
- Clusters where a namespace admin account or token could be stolen or phished.
A single-tenant cluster where only trusted cluster admins touch policies has a lower practical risk, but the fix is still worth applying.
How to check your exposure
First find out which Kyverno version you run. This read-only command lists container images in the kyverno namespace; look for tags below 1.19.2. The namespace name may differ in your install.
kubectl get pods -n kyverno -o jsonpath='{range .items[*]}{.metadata.name}{"\t"}{range .spec.containers[*]}{.image}{" "}{end}{"\n"}{end}'Next, review existing namespaced policies that contain generate rules. This read-only command lists every namespaced Policy in the cluster. Look for policies you do not recognize, especially ones created by namespace-level users.
kubectl get policies.kyverno.io -AOpen any suspicious policy with kubectl get -o yaml and check whether its generate rules point at a namespace other than its own, or at cluster-scoped kinds such as ClusterRoles or ClusterRoleBindings. Then look for resources those rules may have produced that nobody on your team created.
What to do now
The fix is to run Kyverno 1.19.2 or later. Because the advisory text is brief, confirm the exact release notes and upgrade path in the official Kyverno project before changing production.
- Identify every cluster running Kyverno and record its version.
- Upgrade any version below 1.19.2 to 1.19.2 or later after testing.
- List namespaced Policies and review those with generate rules, especially ones targeting other namespaces or cluster-scoped kinds.
- Check for unexpected resources, roles or bindings created by the background controller, and remove only what you confirm is unauthorized.
- Review who holds namespace admin rights and who can create Policy objects; remove access that is not needed.
- If you find signs of abuse, rotate credentials for the affected namespace admin accounts and review the controller's service account permissions.
Until you can upgrade, reduce who can create Policy resources in tenant namespaces. If RBAC allows it, limit that right to trusted administrators. This is a stopgap, not a substitute for the fix.
Detection and monitoring
Kubernetes audit logs are your best source. Watch for creation or update of Policy objects by non-cluster-admin users, and for new cluster-scoped objects created by the Kyverno background controller's service account. Alert on new ClusterRoleBindings in particular, because those are a common route to further control.
If your cluster fronts public websites, Webforta Monitor can watch the public side for certificate, DNS and header changes, though it does not see inside your cluster. For broader background reading, see the Webforta security news and analysis and the security guides hub.
What is still unknown
- Exploitation: the research shows no reports of attacks and no CISA KEV listing, but absence of reports is not proof of safety.
- Exact affected range: only 'before 1.19.2' is stated; the lower bound is not given.
- Independent confirmation: only one primary description exists so far, and NVD repeats it.
Sources
This article is based on the public sources below. Details can change as vendors and researchers publish more; check the original advisories before acting on version numbers.
