Access control with IAM

This page describes access control with Identity and Access Management (IAM) in Artifact Registry.

Default permissions for Artifact Registry minimize setup effort when implementing a CI/CD pipeline. You can also integrate Artifact Registry with third-party CI/CD tools and configure the permissions and authentication required to access repositories.

Before you begin

1. Enable Artifact Registry, including enabling the API and installing the Google Cloud CLI.
2. If you want to apply repository-specific permissions, then create an Artifact Registry repository for your packages.

Overview

IAM permissions and roles determine your ability to create, view, edit, or delete data in an Artifact Registry repository.

A role is a collection of permissions. You can't grant a principal permissions directly; instead, you grant them a role. When you grant a role to a principal, you grant them all the permissions that the role contains. You can grant multiple roles to the same principal.

Google Cloud default permissions

By default, the following permissions apply to Google Cloud CI/CD services in the same project as Artifact Registry:

• Cloud Build permissions include permissions to upload and download artifacts.

• Compute Engine, supported Google Kubernetes Engine versions, and Cloud Run use the Compute Engine default service account, which has read-only access to storage.

If all your services are in the same Google Cloud project and the default permissions meet your needs, you don't need to configure permissions.

You must configure Artifact Registry permissions for these services if:

• You want to use these services to access Artifact Registry in another project. In the project with Artifact Registry, grant the workload identity pool or service account for each service the required role. If connecting to Cloud Run, grant the Cloud Run Service Agent the required role.

• You're using a GKE version that does not have built-in support for pulling images from Artifact Registry. See the GKE section for configuration instructions.
• You want the default service account to have read and write access to repositories. See the following information for details:
  • Compute Engine
  • GKE
• You're using a user-provided service account for your runtime environments instead of the default service account. In the project with Artifact Registry, grant your service account the required role.

Third-party integration

For third-party clients, you must configure both permissions and authentication.

Traditionally, applications running outside Google Cloud use service account keys to access Google Cloud resources. However, service account keys are powerful credentials, and can present a security risk if they are not managed correctly.

Workload Identity Federation lets you use Identity and Access Management to grant external identities IAM roles, including the ability to impersonate service accounts. This approach eliminates the maintenance and security burden associated with service account keys.

Use Workload Identity Federation:

1. Create a Workload Identity Federation pool.
2. Create a Workload Identity Federation provider.
3. Grant the appropriate Artifact Registry role to the workload identity pool to allow repository access.

4. Configure your third-party client to authenticate with Artifact Registry.

   • Container images: Docker, Helm
   • Language packages: Java, Node.js, Python, Go
   • OS packages: Debian, RPM
   • Other: Kubeflow Pipeline templates

Use a service account:

1. Create a service account to act on behalf of your application, or choose an existing service account that use for your CI/CD automation.
2. Grant the appropriate Artifact Registry role to the service account to provide repository access.

3. Configure your third-party client to authenticate with Artifact Registry.

   • Container images: Docker, Helm
   • Language packages: Java, Node.js, Python, Go
   • OS packages: Debian, RPM
   • Other: Kubeflow Pipeline templates

GitLab on Google Cloud

The GitLab on Google Cloud integration uses Workload Identity Federation for authorization and authentication for GitLab workloads on Google Cloud without the need for service accounts or service account keys. For more information on how Workload Identity Federation is used in this partnership, see Google Cloud Workload Identity Federation and IAM policies.

To set up Workload Identity Federation and the necessary IAM roles for the GitLab on Google Cloud, see the GitLab tutorial Google Cloud Workload Identity Federation and IAM policies.

To connect your Artifact Registry repository, follow the GitLab tutorial Google Artifact Registry.

Roles and permissions

Every Artifact Registry API method requires that the principal making the request has the required permissions to use the resource. Permissions are given to principals by setting policies that grant the principal a predefined role on the resource.

You can grant roles on the Google Cloud project or the Artifact Registry repository.

Predefined Artifact Registry roles

IAM provides predefined roles that grant access to specific Google Cloud resources.

Use the following predefined roles for repositories on the pkg.dev domain:

Basic IAM roles

Basic roles are highly permissive roles that existed prior to the introduction of IAM. You shouldn't grant basic roles in a production environment, but you can grant them in a development or test environment.

Use predefined roles for repository access whenever possible so that users and service accounts only have the permissions that are required.

For more information on basic roles, see IAM basic and predefined roles reference.

Granting roles

Grant roles at the project level if the same roles apply to all repositories in the project. If some accounts require different levels of access, grant roles at the repository level.

If you're granting roles using the gcloud command, you can specify a single role binding for a principal or you can make large-scale policy changes by getting a resource's allow policy, modifying it, and then setting the modified allow policy. For more information, see Grant or revoke multiple roles programmatically.

Granting project-wide roles

Grant a role at the project level if the same permissions apply to all repositories in the project.

To add a user or service account to a project and grant them an Artifact Registry role:

Granting repository-specific roles

Grant repository-level roles when you want users or service accounts to have different levels of access for each repository in your project.

provider "google" {
    project = "PROJECT-ID"
}

resource "google_artifact_registry_repository" "my-repo"     {
  provider = google-beta

  location = "LOCATION"
  repository_id = "REPOSITORY"
  description = "DESCRIPTION"
  format = "FORMAT"
}

resource "google_service_account" "repo-account" {
  provider = google-beta

  account_id   = "ACCOUNT-ID"
  display_name = "Repository Service Account"
}

resource "google_artifact_registry_repository_iam_member" "repo-iam" {
  provider = google-beta

  location = google_artifact_registry_repository.my-repo.location
  repository = google_artifact_registry_repository.my-repo.name
  role   = "roles/artifactregistry.reader"
  member = "serviceAccount:${google_service_account.repo-account.email}"
}

Configuring public access to a repository

If you have artifacts that you want to make available to anyone on the internet without authentication, store them in a repository that you make public.

To configure a repository for public read-only access, grant the Artifact Registry Reader role to the principal allUsers. We also recommend capping user request quotas so that a single user can't use up your project's overall quota.

Revoking roles

To revoke access to a repository, remove the principal from the list of authorized principals.

To remove public access from a repository, remove the allUsers principal.

Granting conditional access with tags

Project administrators can create tags for resources across Google Cloud and manage them in Resource Manager. When you attach a tag to a Artifact Registry repository, administrators can use the tag with IAM conditions to grant conditional access to the repository.

You cannot attach tags to individual artifacts.

For more information see the following documentation:

• Administrators setting up tags and access control
  • Creating and managing tags
  • Tags and access control.
• Developers attaching tags to repositories
  • Tagging repositories

Integrating with Google Cloud services

For most Google Cloud service accounts, configuring access to a registry only requires granting the appropriate IAM roles.

Default service accounts for Google Cloud services

Google Cloud services such as Cloud Build or Google Kubernetes Engine use a default service account or service agent to interact with resources within the same project.

You must configure or modify permissions yourself if:

• The Google Cloud service is in a different project than Artifact Registry.
• The default permissions don't meet your needs.
• You're using a user-provided service account to interact with Artifact Registry instead of the default service account.
• Your organizational policy configuration prevents automatic role grants to default service accounts.

The following service accounts typically access Artifact Registry. The email address for the service account includes the Google Cloud project ID or project number of the project where the service is running.

Depending on your organization policy configuration, the default service account might automatically be granted the Editor role on your project. We strongly recommend that you disable the automatic role grant by enforcing the iam.automaticIamGrantsForDefaultServiceAccounts organization policy constraint. If you created your organization after May 3, 2024, this constraint is enforced by default.

If you disable the automatic role grant, you must decide which roles to grant to the default service accounts, and then grant these roles yourself.

If the default service account already has the Editor role, we recommend that you replace the Editor role with less permissive roles. To safely modify the service account's roles, use Policy Simulator to see the impact of the change, and then grant and revoke the appropriate roles.

Granting access to Compute Engine instances

VM instances that access repositories must have both Artifact Registry permissions and storage access scope configured.

While a service account's access level is determined by the IAM roles granted to the service account, access scopes on a VM instance determine the default OAuth scopes for requests made through the gcloud CLI and client libraries on the instance. As a result, access scopes potentially further limit access to API methods when authenticating with Application Default Credentials.

Compute Engine uses the following defaults:

• The Compute Engine default service account is the identity for VM instances. The service account email address has the suffix @developer.gserviceaccount.com.
• The default service account has the IAM basic Editor role, if you have not disabled this behavior.
• Instances you create with the default service account have the Compute Engine default access scopes, including read-only access to storage. While the Editor role generally grants write access, the read-only storage access scope limits the instance service account to downloading artifacts only from any repository in the same project.

You must configure the access scope of the service account if:

• The VM service account needs to access a repository in a different project.
• The VM service account needs to perform actions other than reading artifacts from repositories. This typically applies third-party tools on a VM that need to push images or run Artifact Registry gcloud commands.

To configure roles and set the access scope:

1. In the project with your VM instance, get the name of the Compute Engine default service account. The service account email address has the suffix @developer.gserviceaccount.com.

2. In the project with the repository, grant permissions so that the service account can access the repository.

3. Set the access scope with the --scopes option.

   1. Stop the VM instance. See Stopping an instance.

   2. Set the access scope with the following command:

      gcloud compute instances set-service-account INSTANCE --scopes=SCOPE
      

      Replace SCOPE with the appropriate value.

      • For Docker, the following options are supported:

        • storage-ro - Grants read permission only for pulling images.
        • storage-rw - Grants read and write permission for pushing or pulling images.
        • cloud-platform - View and manage data, including metadata, across Google Cloud service.

      • For other formats, you must use the cloud-platform scope.

   3. Restart the VM instance. See Starting a stopped instance.

Granting access to Google Kubernetes Engine clusters

GKE clusters and node pools can pull containers without any additional configuration if all the following requirements are met:

• GKE is in the same project as Artifact Registry
• Nodes are using the default service account, the Compute Engine default service account
• Nodes were created with read access to storage by:
  • Using the Compute Engine default access scopes.
  • Granting the cloud-platform access scope or another scope that includes read access to storage.
• You're running a supported version of GKE

If your GKE environment does not meet these requirements the instructions to grant access depend on whether you're using the Compute Engine default service account or a user-provided service account as the identity for your nodes.

Default service account

The following configuration requirements apply to the Compute Engine default service account:

1. If GKE is in a different project than Artifact Registry, grant the required permissions to the service account.

2. To push images, interact with repositories for formats other than containers, or run gcloud commands from your cluster, you must set access scopes for the service account when you create the cluster or node pool.

3. If you're not using a supported version of GKE, configure imagePullSecrets.

User-provided service account

If you want to use a user-provided service account as the identity for a cluster, you must:

1. Grant the required permissions to the service account from the Google Cloud project where Artifact Registry is running.

2. By default, creating a cluster or node pool with a user-provided service account grants the cloud-platform access scope.

   If you use the --scopes flag with the gcloud container clusters create or gcloud container node-pools create command, you must include the appropriate access scopes for use with Artifact Registry.

Setting access scopes

Access scopes are the legacy method of specifying authorization for Compute Engine VMs. To pull images from Artifact Registry repositories, GKE nodes must have the storage read-only access scope or another storage access scope that includes storage read access.

You can only set access scopes when you create a cluster or node pool. You cannot change access scopes on existing nodes.

• If you're using the Compute Engine default service account, GKE creates nodes with the Compute Engine default access scopes, which includes read-only access to storage.
• If you're using a user-provided service account, GKE creates nodes with the cloud-platform scope, the scope required for most Google Cloud services.

To specify access scopes when creating a cluster, run the following command:

gcloud container clusters create NAME --scopes=SCOPES

To specify access scopes when creating a node pool, run the following command:

gcloud container node-pools create NAME --scopes=SCOPES

Replace the following values:

• NAME is the name of the cluster or node pool.

• SCOPES is a comma-separated list of access scopes to grant.

  • To access Docker repositories, use one of the following scopes:

  • storage-ro - Grants read-only permission for pulling images.

  • storage-rw - Grants read and write permission for pushing or pulling images.

  • cloud-platform - View and manage data, including metadata, across Google Cloud service.

  • To access other repositories, you must use the cloud-platform scope.

  For a full list of scopes, see the documentation for gcloud container clusters create or gcloud container node-pools create.

For more information about scopes you can set when creating a new cluster, refer to the documentation for the command gcloud container clusters create.

Configuring an imagePullSecret

To configure an imagePullSecret:

1. In the project with GKE, find Compute Engine default service account. The account email address has the suffix @developer.gserviceaccount.com.

2. Download the service account key for the service account.

3. In the project with the repository, verify that you have granted permissions to the repository.

4. In the project with your cluster, create an imagePullSecret secret called artifact-registry with the service account key.

   kubectl create secret docker-registry artifact-registry \
   --docker-server=https://LOCATION-docker.pkg.dev \
   --docker-email=SERVICE-ACCOUNT-EMAIL \
   --docker-username=_json_key \
   --docker-password="$(cat KEY-FILE)"
   

   Where

   • LOCATION is the regional or multi-regional location of the repository.
   • SERVICE-ACCOUNT-EMAIL is the email address of the Compute Engine service account.
   • KEY-FILE is the name of your service account key file. For example key.json.

5. Open your default service account:

   kubectl edit serviceaccount default --namespace default

   Every namespace in your Kubernetes cluster has a default service account called default. This default service account is used to pull your container image.

6. Add the newly created imagePullSecret secret to your default service account:

   imagePullSecrets:
   - name: artifact-registry
   

   Your service account should now look like this:

   apiVersion: v1
   kind: ServiceAccount
   metadata:
     name: default
     namespace: default
     ...
   secrets:
   - name: default-token-zd84v
   # The secret you created:
   imagePullSecrets:
   - name: artifact-registry
   

Now, any new pods created in the current default namespace will have the imagePullSecret secret defined.

Artifact Registry service account

The Artifact Registry Service Agent is a Google-managed service account that acts on behalf of Artifact Registry when interacting with Google Cloud services. For more information about the account and its permissions, see Artifact Registry service account.

What's next

After you have set up permissions, learn more about working with your artifacts.

• Container images: Docker, Helm
• Language packages: Java, Node.js, Python, Go
• OS packages: Debian, RPM