Opscraft
DevSecOps & secrets management · Lesson 4 of 7 · 3 min read

Kubernetes workload and cluster security

Pod security

apiVersion: apps/v1
kind: Deployment
metadata: { name: orders, namespace: orders }
spec:
  template:
    spec:
      serviceAccountName: orders
      automountServiceAccountToken: false     # unless the app calls the K8s API
      securityContext:
        runAsNonRoot: true
        runAsUser: 10001
        fsGroup: 10001
        seccompProfile: { type: RuntimeDefault }
      containers:
        - name: app
          image: 123456789012.dkr.ecr.ap-south-1.amazonaws.com/orders@sha256:abc...
          securityContext:
            allowPrivilegeEscalation: false
            readOnlyRootFilesystem: true
            capabilities: { drop: ["ALL"] }
          resources:
            requests: { cpu: 200m, memory: 512Mi }
            limits: { memory: 512Mi }
          volumeMounts: [{ name: tmp, mountPath: /tmp }]
      volumes: [{ name: tmp, emptyDir: {} }]

Pod Security Admission (built in, replaced PodSecurityPolicy, removed in 1.25) enforces the Pod Security Standards per namespace via labels:

metadata:
  labels:
    pod-security.kubernetes.io/enforce: restricted
    pod-security.kubernetes.io/warn: restricted
    pod-security.kubernetes.io/audit: restricted

Levels: privileged (unrestricted), baseline (blocks known privilege escalations: hostNetwork, privileged, hostPath...), restricted (hardened: non-root, drop ALL capabilities, seccomp, no privilege escalation). Use Kyverno/Gatekeeper for anything more specific (allowed registries, required labels, signature verification, no latest tag).

Enlarge diagramrunAsNonRoot: trueallowPrivilegeEscalation: falsereadOnlyRootFilesystem: truecapabilities: drop [ALL]seccompProfile: RuntimeDefaultautomountServiceAccountToken: falseno root inside the containerno setuid tricksattackers cannot drop filesno raw sockets, no chown…blocks dangerous syscallsno API token if not needed
Figure 1. The settings that harden a pod, and what each stops.
Swipe sideways to see all of it, or tap to open it full screen.

Cluster-level controls

  • RBAC least privilege: namespaced Roles for teams; avoid cluster-admin bindings; no wildcard verbs/resources; restrict secrets read, pods/exec, create pods (which implies the power of any SA in that namespace), escalate/bind/impersonate verbs. Audit with kubectl auth can-i --list --as=..., rbac-tool, Kubescape.
  • EKS access: access entries/IAM Identity Center mappings for humans, private API endpoint or restricted public CIDRs, envelope encryption of Secrets with KMS, control plane audit logs to CloudWatch.
  • Network policies: default-deny ingress (and egress where practical) per namespace, then allow specific flows; on EKS enable network policy in the VPC CNI or use Calico/Cilium. Block pod access to the instance metadata endpoint.
  • Node security: EKS-optimised or Bottlerocket AMIs (minimal, immutable, SELinux), patched regularly via managed node group updates or Karpenter drift; IMDSv2 with hop limit 1; no SSH, SSM if needed.
  • Admission policies: allowed registries (only your ECR), signed images, no privileged, required resources, no hostPath.
  • Runtime detection: Falco rules (shell spawned in container, writes to /etc, unexpected outbound connections) or GuardDuty Runtime Monitoring; feed alerts to the SOC.
  • Benchmarks: CIS Kubernetes/EKS benchmark with kube-bench; Kubescape against NSA/CISA hardening guidance.
Enlarge diagramAccessSSO access entries, private endpointAuthorizationleast-privilege RBAC, no wildcardsAdmissionallowed registries, signed images, no privilegedNetworkdefault-deny NetworkPoliciesNodesBottlerocket / AL2023, patched, no SSHDetectionFalco / GuardDuty Runtime, audit logs
Figure 2. Cluster-level security in layers: who gets in, what they may do, what may run, and what is watched.
Swipe sideways to see all of it, or tap to open it full screen.