On this page (6 sections)
CVE-2026-122304 is a critical authorization bypass in Kyverno, the Kubernetes policy engine. According to the GitHub Advisory Database entry, versions from 1.17.0 before 1.19.2 let a namespace admin step outside their namespace by abusing the cleanup controller.
If you run Kubernetes clusters where different teams or tenants hold admin rights in their own namespaces, this matters to you. The advisory text is short and no exploitation reports appear in the research, so treat this as a patch-soon issue rather than a confirmed active attack.
What happened
The advisory says the Kyverno cleanup controller builds its context loader with a nil policy. In plain terms, a check that should limit what a policy can reach is effectively skipped. A user who is admin of a single namespace can then do things that should be reserved for cluster-level administrators.
The record was published on November 5, 2026 and is also mirrored on NVD. The NVD page repeats the same description, so it is not separate confirmation. The research found only one independent publisher covering the story so far, and the GitHub entry is marked as unreviewed.
Technical details
Kyverno lets users define a CleanupPolicy, which removes resources on a schedule. A policy can include context entries that fetch extra data, such as apiCall or globalReference entries. Per the advisory, an attacker can create a namespaced CleanupPolicy with such entries and have the cleanup controller resolve them using its own ServiceAccount rather than the permissions of the namespace admin.
The advisory states two consequences:
- Reading Secrets in other namespaces, using the cleanup controller's ServiceAccount.
- Modifying validatingwebhookconfigurations, which are cluster-wide objects that control admission checks.
This is a confused deputy problem: a privileged component acts on behalf of a less privileged user without applying that user's limits. Changing validating webhook configurations is especially serious because it can weaken the admission controls that protect the rest of the cluster.
The CVSS v4 metrics listed with the advisory are network attack vector, low complexity, no attack requirements, low privileges required and no user interaction. Confidentiality, integrity and availability impact are rated high for both the vulnerable and the subsequent systems. The score of 9.4 appears in the GitHub entry; the research does not state who assigned it.
Who is affected
You are exposed if both of these are true:
- You run Kyverno at a version from 1.17.0 up to, but not including, 1.19.2.
- Someone you do not fully trust can create CleanupPolicy resources in a namespace, for example a tenant team with namespace admin rights.
Single-team clusters where only cluster admins hold admin rights have a smaller practical risk, since the attacker needs at least low privileges to start. Multi-tenant platforms, shared clusters and CI environments that hand out namespace admin are the realistic targets.
How to check your exposure
This read-only command lists the container images of pods across all namespaces and filters for Kyverno, so you can see the running version tag. Run it with kubectl against each cluster; look for tags from 1.17.0 to 1.19.1.
kubectl get pods -A -o jsonpath='{range .items[*]}{.metadata.namespace}{"\t"}{.metadata.name}{"\t"}{range .spec.containers[*]}{.image}{" "}{end}{"\n"}{end}' | grep -i kyvernoIf you installed with Helm, this read-only command shows the releases and chart versions, filtered for Kyverno. Chart versions and application versions can differ, so confirm the image tag as well.
helm list -A | grep -i kyvernoNext, review which CleanupPolicy objects exist. This read-only command lists namespaced CleanupPolicy resources in every namespace; unexpected entries, especially ones with apiCall or globalReference context, deserve a closer look.
kubectl get cleanuppolicies -AWhat to do now
The affected range ends before 1.19.2, so the working assumption is that Kyverno 1.19.2 or later is no longer in the affected range. The research does not include a vendor statement naming a patched version, so confirm it in the Kyverno advisories linked from the record: GHSA-p58h-r4j3-h5vx, GHSA-rcxg-3f92-3mgr and the VulnCheck advisory. I did not review the contents of those pages, so check them for fix and workaround details.
- Identify every cluster running Kyverno and record the version.
- Confirm the fixed version in the linked Kyverno and VulnCheck advisories.
- Upgrade any instance from 1.17.0 to 1.19.1 to 1.19.2 or later, following your normal upgrade method.
- List existing CleanupPolicy objects and review any that use apiCall or globalReference context entries.
- Review who holds namespace admin rights and remove access that is not needed.
- Check validatingwebhookconfigurations for entries you do not recognize.
- Consider rotating Secrets that the cleanup controller's ServiceAccount could read, if you suspect misuse.
Detection and monitoring
The sources give no indicators of compromise, so detection is about reviewing change. Kubernetes audit logs, if enabled, are the best record. Look for creation of CleanupPolicy objects by namespace-level users, and for updates to validatingwebhookconfigurations by the cleanup controller's ServiceAccount, which would be unusual.
Also watch for cross-namespace Secret reads by that ServiceAccount. Keep alerting on admission webhook changes in general, since they alter how your whole cluster enforces policy.
This issue sits inside the cluster, so outside checks have limited reach. For the public side of your stack, you can still keep an eye on certificates, DNS and uptime with Webforta Monitor. For broader habits around access and updates, see small business website security.
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.
