Securing access to resources with IAM

Kubernetes users and service accounts need permissions to manage Config Connector resources. With Config Connector, your project's control plane can be managed by identities that use Kubernetes Role-Based Access Control (RBAC). You can also reference Identity and Access Management (IAM) Policies.

Resources that can reference IAMPolicy and IAMPolicyMember are listed in the Resource reference. These resources have the property "Can Be Referenced by IAMPolicy/IAMPolicyMember".

This topic explains how to secure access to Google Cloud resources using Identity and Access Management.

Before you begin

Install Config Connector on your cluster.

Securing control plane access with RBAC

In this example, you will create a service account and grant it permissions to manage a PubSubTopic. This service account cannot manage other types of Config Connector resources.

  1. Create a file named pubsub-topic-service-account.yaml with the following contents:

    apiVersion: v1
    kind: ServiceAccount
    metadata:
      name: pubsub-topic-service-account
      namespace: default
    

    Apply this to create the pubsub-topic-service-account service account:

    kubectl apply -f pubsub-topic-service-account.yaml --namespace CC_NAMESPACE

    Replace CC_NAMESPACE with the namespace Config Connector manages resources from.

  2. Confirm pubsub-topic-service-account cannot create PubSubTopic resources by verifying the output of the following command contains no:

    kubectl auth can-i get pubsubtopics --as=system:serviceaccount:default:pubsub-topic-service-account
  3. Next, create a ClusterRole that allows Pub/Sub topic creation.

    The ClusterRole can only manage resources that have values specified in rules.apiGroups and rules.resources. To find values for apiGroups and resources, see the reference for your resources.

    Create a file named pubsub-topic-editor-role.yaml with the following contents:

    apiVersion: rbac.authorization.k8s.io/v1
    kind: ClusterRole
    metadata:
      creationTimestamp: null
      name: pubsub-topic-editor
    rules:
    - apiGroups:
      - pubsub.cnrm.cloud.google.com
      resources:
      - pubsubtopics
      verbs:
      - get
      - list
      - watch
      - create
      - update
      - patch
      - delete
    

    Apply pubsub-topic-editor.yaml to create the ClusterRole:

    kubectl apply -f pubsub-topic-editor-role.yaml --namespace CC_NAMESPACE

    Replace CC_NAMESPACE with the namespace Config Connector manages resources from.

  4. Next, create a RoleBinding between the ClusterRole and your service account. Create a file named pubsub-topic-editor-rolebinding.yaml with the following contents:

    apiVersion: rbac.authorization.k8s.io/v1
    kind: RoleBinding
    metadata:
      name: pubsub-topic-editor-rolebinding.
    subjects:
    - kind: ServiceAccount
      name: pubsub-topic-service-account
    roleRef:
      apiGroup: rbac.authorization.k8s.io
      kind: ClusterRole
      name: pubsub-topic-editor
    
  5. Apply pubsub-topic-editor-rolebinding.yaml to your cluster.

    kubectl apply -f pubsub-topic-editor-rolebinding.yaml --namespace CC_NAMESPACE

    Replace CC_NAMESPACE with the namespace Config Connector manages resources from.

  6. Confirm the pubsub-topic-service-account is allowed to create PubSubTopic resources by confirming the output of the following command is yes:

    kubectl auth can-i get pubsubtopics \
      --as=system:serviceaccount:default:pubsub-topic-service-account

Cleaning up

Use kubectl delete to remove the Service Account, IAM Role and Rolebinding.

kubectl delete -f pubsub-topic-editor-rolebinding.yaml --namespace CC_NAMESPACE
kubectl delete -f pubsub-topic-editor-role.yaml --namespace CC_NAMESPACE
kubectl delete -f pubsub-topic-service-account.yaml --namespace CC_NAMESPACE

Replace CC_NAMESPACE with the namespace Config Connector manages resources from.

Securing the data plane with IAM Policies

In this example, you use the permissions granted earlier to create a PubSubTopic and limit access to it with an IAMPolicyMember resource.

  1. Create a file named pubsub-topic-sample.yaml with the following content:

    apiVersion: pubsub.cnrm.cloud.google.com/v1beta1
    kind: PubSubTopic
    metadata:
      name: pubsubtopic-sample
    

    Apply pubsub-topic-sample.yaml with kubectl:

    kubectl apply -f pubsub-topic-sample.yaml --namespace CC_NAMESPACE

    Replace CC_NAMESPACE with the namespace Config Connector manages resources from.

  2. Create a file named iampolicymember.yaml with the following content, replacing EMAIL_ADDRESS with your Google Cloud account's email address:

    apiVersion: iam.cnrm.cloud.google.com/v1beta1
    kind: IAMPolicyMember
    metadata:
      name: iampolicymember-sample
    spec:
      resourceRef:
        apiVersion: pubsub.cnrm.cloud.google.com/v1beta1
        kind: PubSubTopic
        name: pubsubtopic-sample
      role: roles/pubsub.admin
      member: "user:EMAIL_ADDRESS"
    
  3. Apply the iampolicymember.yaml.

    kubectl apply -f iampolicymember.yaml  --namespace CC_NAMESPACE 

    Replace CC_NAMESPACE with the namespace Config Connector manages resources from.

  4. Confirm the policy has been applied to Google Cloud by running this command and looking for your email address in the output, replacing PROJECT_ID with your project ID:

    gcloud beta pubsub topics get-iam-policy projects/PROJECT_ID/topics/pubsubtopic-sample

Access to your Pub/Sub topics is now protected with an IAMPolicyMember.

Cleaning up

Use kubectl delete to remove the Pub/Sub topic and IAMPolicyMember from your Google Cloud Project.

kubectl delete -f iampolicymember.yaml --namespace CC_NAMESPACE
kubectl delete -f pubsub-topic-sample.yaml --namespace CC_NAMESPACE

Security best practices for multi-tenant isolation

To help prevent unauthorized privilege escalation and enforce tenant isolation, follow these best practices:

  • Enforce namespaced mode: avoid cluster mode in multi-tenant environments. In cluster mode, all namespaces share a single IAM service account. Any user who can create Config Connector resources in any namespace can act with the full IAM permissions of that service account.

  • Restrict RBAC for IAM resources: limit creation and modification permissions for IAMPolicy, IAMPartialPolicy, and IAMPolicyMember resources by using Kubernetes role-based access control (RBAC). If developers have broad write access to these resources, they can alter Google Cloud IAM bindings.

  • Scope IAM bindings for Workload Identity Federation for GKE: configure these bindings strictly for designated Kubernetes ServiceAccounts in specific namespaces. In namespaced mode, bind each IAM service account only to the Config Connector Kubernetes ServiceAccount for its namespace:

    gcloud iam service-accounts add-iam-policy-binding \
        GSA_NAME@PROJECT_ID.iam.gserviceaccount.com \
        --member="serviceAccount:PROJECT_ID.svc.id.goog[cnrm-system/cnrm-controller-manager-CC_NAMESPACE]" \
        --role="roles/iam.workloadIdentityUser"
    

    Replace the following:

    • GSA_NAME: the name of the IAM service account that Config Connector uses for the namespace.
    • PROJECT_ID: the ID of the project that contains your cluster and the IAM service account.
    • CC_NAMESPACE: the namespace that Config Connector manages resources from.

    Don't grant the roles/iam.workloadIdentityUser role to a principal set, such as all Pods in the cnrm-system namespace. In namespaced mode, the Config Connector controllers for all namespaces run in cnrm-system, so this type of binding lets every namespace's controller impersonate the IAM service account.

What's next

Use Secrets to pass information securely to Google Cloud resources.