Lack of Cluster Level Policy Enforcement
Move the tag, and production pulls it on its own.
What Is Lack of Cluster Level Policy Enforcement?
A deployment that names a mutable tag runs whatever that tag points to when each pod starts, and a cluster with no enforced admission policy never asks. You'll push a changed image under the same tag and watch rescheduled production pods pull it with no deploy. Then you'll find the build that is really running, pin the workload by digest and switch an audit-only ValidatingAdmissionPolicy to Deny so the cluster rejects tags. Maps to OWASP Kubernetes Top Ten K04, Lack of Cluster Level Policy Enforcement.
What You'll Learn in Lack of Cluster Level Policy Enforcement
- Explain why a mutable tag lets a different image reach production without a deploy.
- Find the digest each pod is actually running from its container status.
- Pin workloads by digest and enforce it with a ValidatingAdmissionPolicy.
- Distinguish a policy in Audit mode from one that denies requests.
Lack of Cluster Level Policy Enforcement — Training Steps
-
A tag, not an image
Bob has a registry push token for Ostravel's booking images, leaked in a build log. He has no access to the cluster and does not need any. Ostravel publishes its deployment manifests in an open repository, and the fares service's manifest tells him how production chooses its image. Ostravel sells flights and hotels. The fares service handles every search and booking.
-
Whatever latest is today
Two lines decide that production will run whatever the tag points to at the moment a pod starts.
-
Move the tag
Bob builds an image from the public fares-api code with one addition: it copies every booking it handles to a server he controls. He pushes it under the same tag.
-
Production pulls it
Within the hour, Kubernetes reschedules fares-api pods as it always does. Each new pod pulls latest. Bob checks his collection server.
-
Check your understanding
-
Two builds behind one tag
Ostravel's image inventory finds fares-api pods running two different digests under the same tag, one of which no pipeline built.
-
See what is actually running
A tag says nothing about which build a pod runs. The image ID on each pod's status does.
-
Open the deployment and the policy
Alice opens the fares-api deployment and the cluster's image policy.
-
Pin the build by digest
Alice replaces the tag with the digest of CI build 1184, the last build the pipeline produced and signed off. A digest names exactly one image and cannot be moved.
-
A policy that only watched
The cluster already has a policy that requires digests. Someone wrote it and bound it, but in audit mode, so it recorded every violation and blocked none.
Security Framework Coverage
OWASP Kubernetes Top 10
- K04:2025 Lack of Cluster Level Policy Enforcement
CWE
- CWE-345 Insufficient Verification of Data Authenticity
- CWE-693 Protection Mechanism Failure
MITRE ATT&CK
- T1525 Implant Internal Image
- T1610 Deploy Container
CIS Controls
- CIS 4 Secure Configuration of Enterprise Assets and Software
NIST CSF
- PR.AT-02 Individuals in specialized roles are provided with awareness and training so that they possess the knowledge and skills to perform relevant tasks with cybersecurity risks in mind
- PR.PS Platform Security
NICE Framework
- IO-WRL-005 Systems Administration
- DD-WRL-001 Cybersecurity Architecture