Reviewing and applying recommendations

This page explains how to view, understand, and apply recommendations made by the IAM recommender. The IAM recommender helps you enforce the principle of least privilege by ensuring that members have only the permissions that they actually need.

Before you begin

• Understand the IAM recommender.
• Make sure you have the IAM permissions that you need in order to work with the IAM recommender.
• Review the IAM recommender best practices.

Reviewing and applying recommendations

The easiest way to review and apply your recommendations is to use the Google Cloud Console. You must use the Cloud Console to complete some tasks:

• Viewing details about why you received a recommendation
• Dismissing recommendations
• Automatically creating a custom role when you apply a recommendation

You can also review and apply recommendations with the gcloud command-line tool and the Recommender API.

Note: The IAM recommender does not take other forms of access controls into account when making recommendations. If you use access controls that are separate from IAM, such as Cloud Storage access control lists (ACLs) or Kubernetes role-based access control (RBAC), ensure that each recommendation will not disrupt those access controls before you apply the recommendation.

Console

1. In the Cloud Console, go to the IAM page.

   Go to the IAM page

2. In the list of members of your project, find the Analyzed permissions column.

   For each role granted to a member, this column shows the number of excess permissions, followed by the total number of permissions in the role:

   Note: The IAM recommender does not analyze permissions for all roles. To learn which roles the IAM recommender analyzes, see How the IAM recommender works.

   

   

   The Recommendation available icon indicates that a recommendation is available for a role, either to revoke the role or to replace it with a role that includes fewer permissions.

3. If there are recommendations to review, click a Recommendation available icon to get details about the recommendation.

   If the recommendation is to replace the role, the IAM recommender always suggests a set of predefined roles that you can apply.

   In some cases, the IAM recommender also suggests creating a new custom role at the project level. If a custom role recommendation is available, the Cloud Console shows it by default. To switch to the predefined role recommendation, click View recommended predefined role.

4. Review the recommendation carefully, and make sure you understand how it will change the member's access to Google Cloud resources. See Reviewing recommendations to understand the types of changes that a recommendation can include.

   Note: A recommendation will never increase a member's level of access. See How the IAM recommender works for more information.

5. (Optional) If the recommendation is to create a custom role, update the Title, Description, ID, and Role launch stage as needed.

   If you need to add permissions to the custom role, click Add permissions.

   If you need to remove permissions from the custom role, clear the checkbox for each permission that you want to remove.

6. Take action on the recommendation.

   To apply the recommendation, click Apply or Create and apply. If you change your mind in the next 90 days, use the IAM recommender history to revert your choice.

   To dismiss the recommendation, click Dismiss, then confirm your choice. You can restore a dismissed recommendation as long as the recommendation is still valid.

7. Repeat the previous steps until you have reviewed all of your recommendations.

gcloud

Review your recommendations:

To list your recommendations, run the gcloud recommender recommendations list command:

gcloud recommender recommendations list \
    --location=global \
    --recommender=google.iam.policy.Recommender \
    --project=project-id \
    --format=format

Replace the following values:

• project-id: The project identifier, such as project-123.
• format: The format of the response. Use json or yaml.

The response is similar to the following example. In this example, a service account has not used any permissions from the Compute Admin role (roles/compute.admin) in the past 90 days. As a result, the IAM recommender suggests that you revoke the role:

[
  {
    "associatedInsights": [
      {
        "insight": "projects/123456789012/locations/global/insightTypes/google.iam.policy.Insight/insights/279ef748-408f-44db-9a4a-1ff8865b9839"
      }
    ],
    "content": {
      "operationGroups": [
        {
          "operations": [
            {
              "action": "remove",
              "path": "/iamPolicy/bindings/*/members/*",
              "pathFilter": {
                "/iamPolicy/bindings/*/condition/expression": "",
                "/iamPolicy/bindings/*/members/*": "serviceAccount:[e-mail]",
                "/iamPolicy/bindings/*/role": "roles/compute.admin"
              },
              "resource": "//cloudresourcemanager.googleapis.com/projects/example-project",
              "resourceType": "cloudresourcemanager.googleapis.com/Project"
            }
          ]
        }
      ]
    },
    "description": "This role has not been used during the observation window.",
    "recommenderSubtype": "REMOVE_ROLE",
    "etag": "\"770237e2c0decf40\"",
    "lastRefreshTime": "2020-01-09T06:06:17Z",
    "name": "projects/123456789012/locations/global/recommenders/google.iam.policy.Recommender/recommendations/fb927dc1-9695-4436-0000-f0f285007c0f",
    "primaryImpact": {
      "category": "SECURITY",
        "securityProjection": {
          "revokedIamPermissionsCount": 708
        }
    },
    "stateInfo": {
      "state": "ACTIVE"
    }
  }
]

Review each recommendation carefully, and consider how it will change the member's access to Google Cloud resources.

To see the permission usage that this recommendation is based on, view the insights that are associated with the recommendation. These insights are listed in the associatedInsights field. To view an insight, do the following:

• Copy the insight ID. The insight ID is everything after insights/ in the insight field. In the preceding example, the insight ID is 279ef748-408f-44db-9a4a-1ff8865b9839.
• Follow the instructions to get an insight, using the insight ID you copied and the insight type ID google.iam.policy.Insight.

To apply a recommendation:

1. Use the gcloud recommender recommendations mark-claimed command to change the recommendation's state to CLAIMED, which prevents the recommendation from changing while you apply it:

   gcloud recommender recommendations mark-claimed \
       recommendation-id \
       --location=global \
       --recommender=google.iam.policy.Recommender \
       --project=project-id \
       --format=format \
       --etag=etag \
       --state-metadata=state-metadata
   

   Replace the following values:

   • recommendation-id: The unique identifier for the recommendation. This value appears at the end of the name field in the recommendation. In the example shown above, the ID is fb927dc1-9695-4436-0000-f0f285007c0f.
   • project-id: The project identifier, such as project-123.
   • format: The format of the response. Use json or yaml.
   • etag: The value of the etag field in the recommendation, such as "dd0686e7136a4cbb". Note that this value can include quotes.
   • state-metadata: Optional. Comma-separated key-value pairs that contain your choice of metadata about the recommendation. For example, --state-metadata=reviewedBy=alice,priority=high. The metadata replaces the stateInfo.stateMetadata field in the recommendation.

   If the command succeeds, the response shows the recommendation in a CLAIMED state, as shown in the following example. For clarity, the example omits most fields:

   [
     {
       "description": "This role has not been used during the observation window.",
       "recommenderSubtype": "REMOVE_ROLE",
       "etag": "\"df7308cca9719dcc\"",
       "name": "projects/123456789012/locations/global/recommenders/google.iam.policy.Recommender/recommendations/fb927dc1-9695-4436-0000-f0f285007c0f",
       "stateInfo": {
         "state": "CLAIMED",
         "stateMetadata": {
           "reviewedBy": "alice",
           "priority": "high"
         }
       }
     }
   ]

2. Get the IAM policy for the project, then modify the policy so that it reflects the recommendation.

3. Update the recommendation's state to SUCCEEDED, if you were able to apply the recommendation, or FAILED, if you could not apply the recommendation:

   gcloud recommender recommendations command \
       recommendation-id \
       --location=global \
       --recommender=google.iam.policy.Recommender \
       --project=project-id \
       --format=format \
       --etag=etag \
       --state-metadata=state-metadata
   

   Replace the following values:

   • command: Use mark-succeeded, if you were able to apply the recommendation, or mark-failed, if you could not apply the recommendation.
   • recommendation-id: The unique identifier for the recommendation. This value appears at the end of the name field in the recommendation. In the example shown above, the ID is fb927dc1-9695-4436-0000-f0f285007c0f.
   • project-id: The project identifier, such as project-123.
   • format: The format of the response. Use json or yaml.
   • etag: The value of the etag field in the recommendation, such as "dd0686e7136a4cbb". Note that this value can include quotes.
   • state-metadata: Optional. Comma-separated key-value pairs that contain your choice of metadata about the recommendation. For example, --state-metadata=reviewedBy=alice,priority=high. The metadata replaces the stateInfo.stateMetadata field in the recommendation.

   For example, if you marked the recommendation as having succeeded, the response shows the recommendation in a SUCCEEDED state. For clarity, this example omits most fields:

   [
     {
       "description": "This role has not been used during the observation window.",
       "recommenderSubtype": "REMOVE_ROLE",
       "etag": "\"dd0686e7136a4cbb\"",
       "name": "projects/123456789012/locations/global/recommenders/google.iam.policy.Recommender/recommendations/fb927dc1-9695-4436-0000-f0f285007c0f",
       "stateInfo": {
         "state": "SUCCEEDED",
         "stateMetadata": {
           "reviewedBy": "alice",
           "priority": "high"
         }
       }
     }
   ]

Note: The IAM recommender also supports insights. IAM policy insights are ML-based findings about permission usage within your project that may or may not be connected to a recommendation. To learn how to use insights, see Managing insights.

REST

Note: These instructions assume that you have authenticated and set the GOOGLE_APPLICATION_CREDENTIALS environment variable.

Review your recommendations:

The Recommender API's recommendations.list method lists all available recommendations for your project.

Before using any of the request data below, make the following replacements:

• project-id: Your Google Cloud project ID.
• page-size: Optional. The maximum number of results to return from this request. If not specified, the server will determine the number of results to return. If the number of recommendations is greater than the page size, the response contains a pagination token that you can use to retrieve the next page of results.
• page-token: Optional. The pagination token returned in an earlier response from this method. If specified, the list of recommendations will start where the previous request ended.
• filter: Optional. A filter expression to restrict the recommendations returned. You can filter recommendations based on the stateInfo.state field. For example, stateInfo.state:"DISMISSED" or stateInfo.state:"FAILED".

HTTP method and URL:

GET https://recommender.googleapis.com/v1/projects/project-id/locations/global/recommenders/google.iam.policy.Recommender/recommendations?pageSize=page-size&pageToken=page-token&filter=filter

To send your request, expand one of these options:

curl (Linux, macOS, or Cloud Shell)

Note: Ensure you have set the GOOGLE_APPLICATION_CREDENTIALS environment variable to your service account private key file path.

Execute the following command:

curl -X GET \
-H "Authorization: Bearer "$(gcloud auth application-default print-access-token) \
https://recommender.googleapis.com/v1/projects/project-id/locations/global/recommenders/google.iam.policy.Recommender/recommendations?pageSize=page-size&pageToken=page-token&filter=filter

PowerShell (Windows)

Note: Ensure you have set the GOOGLE_APPLICATION_CREDENTIALS environment variable to your service account private key file path.

Execute the following command:

$cred = gcloud auth application-default print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
  -Method GET `
  -Headers $headers `
  -Uri "https://recommender.googleapis.com/v1/projects/project-id/locations/global/recommenders/google.iam.policy.Recommender/recommendations?pageSize=page-size&pageToken=page-token&filter=filter" | Select-Object -Expand Content

The response is similar to the following example. In this example, a service account has not used any permissions from the Compute Admin role (roles/compute.admin) in the past 90 days. As a result, the IAM recommender suggests that you revoke the role:

{
  "recommendations": [
    "name": "projects/123456789012/locations/global/recommenders/google.iam.policy.Recommender/recommendations/fb927dc1-9695-4436-0000-f0f285007c0f",
    "description": "This role has not been used during the observation window.",
    "lastRefreshTime": "2020-01-09T06:06:17Z",
    "primaryImpact": {
      "category": "SECURITY",
        "securityProjection": {
          "revokedIamPermissionsCount": 708
        }
    },
    "content": {
      "operationGroups": [
        {
          "operations": [
            {
              "action": "remove",
              "path": "/iamPolicy/bindings/*/members/*",
              "pathFilter": {
                "/iamPolicy/bindings/*/condition/expression": "",
                "/iamPolicy/bindings/*/members/*": "serviceAccount:[e-mail]",
                "/iamPolicy/bindings/*/role": "roles/compute.admin"
              },
              "resource": "//cloudresourcemanager.googleapis.com/projects/example-project",
              "resourceType": "cloudresourcemanager.googleapis.com/Project"
            }
          ]
        }
      ]
    },
    "stateInfo": {
      "state": "ACTIVE"
    }
    "etag": "\"770237e2c0decf40\"",
    "recommenderSubtype": "REMOVE_ROLE",
    "associatedInsights": [
      {
        "insight": "projects/123456789012/locations/global/insightTypes/google.iam.policy.Insight/insights/279ef748-408f-44db-9a4a-1ff8865b9839"
      }
  ]
}

Review each recommendation carefully, and consider how it will change the member's access to Google Cloud resources.

To see the permission usage that this recommendation is based on, view the insights that are associated with the recommendation. These insights are listed in the associatedInsights field. To view an insight, do the following:

• Copy the insight ID. The insight ID is everything after insights/ in the insight field. In the preceding example, the insight ID is 279ef748-408f-44db-9a4a-1ff8865b9839.
• Follow the instructions to get an insight, using the insight ID you copied and the insight type ID google.iam.policy.Insight.

To apply a recommendation:

1. Mark the recommendation as CLAIMED:

   The Recommender API's recommendations.markClaimed method marks a recommendation as CLAIMED, which prevents the recommendation from changing while you apply it.

   Before using any of the request data below, make the following replacements:

   • project-id: Your Google Cloud project ID.
   • recommendation-id: The unique identifier for the recommendation. This value appears at the end of the name field in the recommendation. For example, if the name field is projects/example-project/locations/global/recommenders/google.iam.policy.Recommender/recommendations/fb927dc1-9695-4436-0000-f0f285007c0f, the recommendation ID is fb927dc1-9695-4436-0000-f0f285007c0f.
   • etag: The value of the etag field in the recommendation, such as "dd0686e7136a4cbb". Use backslashes to escape quotes, for example, "\"df7308cca9719dcc\"".
   • state-metadata: Optional. An object that contains key-value pairs with your choice of metadata about the recommendation. For example, {"reviewedBy": "alice", "priority": "high"}. The metadata replaces the stateInfo.stateMetadata field in the recommendation.

   HTTP method and URL:

   POST https://recommender.googleapis.com/v1/projects/project-id/locations/global/recommenders/google.iam.policy.Recommender/recommendations/recommendation-id:markClaimed

   Request JSON body:

   {
     "etag": "etag"
     "stateMetadata": {
       "state-metadata"
   }
   

   To send your request, expand one of these options:

   curl (Linux, macOS, or Cloud Shell)

   Note: Ensure you have set the GOOGLE_APPLICATION_CREDENTIALS environment variable to your service account private key file path.

   Save the request body in a file called request.json, and execute the following command:

   curl -X POST \
   -H "Authorization: Bearer "$(gcloud auth application-default print-access-token) \
   -H "Content-Type: application/json; charset=utf-8" \
   -d @request.json \
   https://recommender.googleapis.com/v1/projects/project-id/locations/global/recommenders/google.iam.policy.Recommender/recommendations/recommendation-id:markClaimed

   PowerShell (Windows)

   Note: Ensure you have set the GOOGLE_APPLICATION_CREDENTIALS environment variable to your service account private key file path.

   Save the request body in a file called request.json, and execute the following command:

   $cred = gcloud auth application-default print-access-token
   $headers = @{ "Authorization" = "Bearer $cred" }
   
   Invoke-WebRequest `
     -Method POST `
     -Headers $headers `
     -ContentType: "application/json; charset=utf-8" `
     -InFile request.json `
     -Uri "https://recommender.googleapis.com/v1/projects/project-id/locations/global/recommenders/google.iam.policy.Recommender/recommendations/recommendation-id:markClaimed" | Select-Object -Expand Content

   The response shows the recommendation in a CLAIMED state, as shown in the following example. For clarity, this example omits most fields:

   {
     "description": "This role has not been used during the observation window.",
     "stateInfo": {
       "state": "CLAIMED",
       "stateMetadata": {
         "reviewedBy": "alice",
         "priority": "high"
       }
     }
     "etag": "\"dd0686e7136a4cbb\"",
     "recommenderSubtype": "REMOVE_ROLE"
   }
   

2. Get the IAM policy for the project, then modify the policy so that it reflects the recommendation.

3. Update the recommendation's state to SUCCEEDED, if you were able to apply the recommendation, or FAILED, if you could not apply the recommendation:

   SUCCEEDED

   The Recommender API's recommendations.markSucceeded method marks a recommendation as SUCCEEDED, indicating that you were able to apply it.

   Before using any of the request data below, make the following replacements:

   • project-id: Your Google Cloud project ID.
   • recommendation-id: The unique identifier for the recommendation. This value appears at the end of the name field in the recommendation. For example, if the name field is projects/example-project/locations/global/recommenders/google.iam.policy.Recommender/recommendations/fb927dc1-9695-4436-0000-f0f285007c0f, the recommendation ID is fb927dc1-9695-4436-0000-f0f285007c0f.
   • etag: The value of the etag field in the recommendation, such as "dd0686e7136a4cbb". Use backslashes to escape quotes, for example, "\"df7308cca9719dcc\"".
   • state-metadata: Optional. An object that contains key-value pairs with your choice of metadata about the recommendation. For example, {"reviewedBy": "alice", "priority": "high"}. The metadata replaces the stateInfo.stateMetadata field in the recommendation.

   HTTP method and URL:

   POST https://recommender.googleapis.com/v1/projects/project-id/locations/global/recommenders/google.iam.policy.Recommender/recommendations/recommendation-id:markSucceeded

   Request JSON body:

   {
     "etag": "etag"
     "stateMetadata": {
       "state-metadata"
   }
   

   To send your request, expand one of these options:

   curl (Linux, macOS, or Cloud Shell)

   Note: Ensure you have set the GOOGLE_APPLICATION_CREDENTIALS environment variable to your service account private key file path.

   Save the request body in a file called request.json, and execute the following command:

   curl -X POST \
   -H "Authorization: Bearer "$(gcloud auth application-default print-access-token) \
   -H "Content-Type: application/json; charset=utf-8" \
   -d @request.json \
   https://recommender.googleapis.com/v1/projects/project-id/locations/global/recommenders/google.iam.policy.Recommender/recommendations/recommendation-id:markSucceeded

   PowerShell (Windows)

   Note: Ensure you have set the GOOGLE_APPLICATION_CREDENTIALS environment variable to your service account private key file path.

   Save the request body in a file called request.json, and execute the following command:

   $cred = gcloud auth application-default print-access-token
   $headers = @{ "Authorization" = "Bearer $cred" }
   
   Invoke-WebRequest `
     -Method POST `
     -Headers $headers `
     -ContentType: "application/json; charset=utf-8" `
     -InFile request.json `
     -Uri "https://recommender.googleapis.com/v1/projects/project-id/locations/global/recommenders/google.iam.policy.Recommender/recommendations/recommendation-id:markSucceeded" | Select-Object -Expand Content

   The response shows the recommendation in a SUCCEEDED state, as shown in the following example. For clarity, this example omits most fields:

   {
     "description": "This role has not been used during the observation window.",
     "stateInfo": {
       "state": "SUCCEEDED",
       "stateMetadata": {
         "reviewedBy": "alice",
         "priority": "high"
       }
     }
     "etag": "\"dd0686e7136a4cbb\"",
     "recommenderSubtype": "REMOVE_ROLE"
   }
   

   FAILED

   The Recommender API's recommendations.markFailed method marks a recommendation as FAILED, indicating that you were not able to apply it.

   Before using any of the request data below, make the following replacements:

   • project-id: Your Google Cloud project ID.
   • recommendation-id: The unique identifier for the recommendation. This value appears at the end of the name field in the recommendation. For example, if the name field is projects/example-project/locations/global/recommenders/google.iam.policy.Recommender/recommendations/fb927dc1-9695-4436-0000-f0f285007c0f, the recommendation ID is fb927dc1-9695-4436-0000-f0f285007c0f.
   • etag: The value of the etag field in the recommendation, such as "dd0686e7136a4cbb". Use backslashes to escape quotes, for example, "\"df7308cca9719dcc\"".
   • state-metadata: Optional. An object that contains key-value pairs with your choice of metadata about the recommendation. For example, {"reviewedBy": "alice", "priority": "high"}. The metadata replaces the stateInfo.stateMetadata field in the recommendation.

   HTTP method and URL:

   POST https://recommender.googleapis.com/v1/projects/project-id/locations/global/recommenders/google.iam.policy.Recommender/recommendations/recommendation-id:markFailed

   Request JSON body:

   {
     "etag": "etag"
     "stateMetadata": {
       "state-metadata"
   }
   

   To send your request, expand one of these options:

   curl (Linux, macOS, or Cloud Shell)

   Note: Ensure you have set the GOOGLE_APPLICATION_CREDENTIALS environment variable to your service account private key file path.

   Save the request body in a file called request.json, and execute the following command:

   curl -X POST \
   -H "Authorization: Bearer "$(gcloud auth application-default print-access-token) \
   -H "Content-Type: application/json; charset=utf-8" \
   -d @request.json \
   https://recommender.googleapis.com/v1/projects/project-id/locations/global/recommenders/google.iam.policy.Recommender/recommendations/recommendation-id:markFailed

   PowerShell (Windows)

   Note: Ensure you have set the GOOGLE_APPLICATION_CREDENTIALS environment variable to your service account private key file path.

   Save the request body in a file called request.json, and execute the following command:

   $cred = gcloud auth application-default print-access-token
   $headers = @{ "Authorization" = "Bearer $cred" }
   
   Invoke-WebRequest `
     -Method POST `
     -Headers $headers `
     -ContentType: "application/json; charset=utf-8" `
     -InFile request.json `
     -Uri "https://recommender.googleapis.com/v1/projects/project-id/locations/global/recommenders/google.iam.policy.Recommender/recommendations/recommendation-id:markFailed" | Select-Object -Expand Content

   The response shows the recommendation in a FAILED state, as shown in the following example. For clarity, this example omits most fields:

   {
     "description": "This role has not been used during the observation window.",
     "stateInfo": {
       "state": "FAILED",
       "stateMetadata": {
         "reviewedBy": "alice",
         "priority": "high"
       }
     }
     "etag": "\"dd0686e7136a4cbb\"",
     "recommenderSubtype": "REMOVE_ROLE"
   }
   

Note: The IAM recommender also supports insights. IAM insights are ML-based findings about permission usage within your project that may or may not be connected to a recommendation. To learn how to use insights, see IAM insights.

Viewing, reverting, and restoring changes

After you apply or dismiss a recommendation, that action appears in the recommendations history. To view the recommendations history:

1. In the Cloud Console, go to the IAM page.

   Go to the IAM page

2. Near the top of the screen, click Recommendations history.

   The Cloud Console shows a list of previous actions on your IAM recommendations.

3. To view details about a recommendation, click the expander arrow.

   The Cloud Console shows details about the action that was taken, including the member that took the action:

   

   

4. (Optional) If necessary, you can revert the recommendation, which undoes the changes in the recommendation, or restore a recommendation that you dismissed.

   To revert a previously applied change for a recommendation, click Revert. The Cloud Console reverts the changes to the member's roles. The recommendation no longer appears in the Cloud Console.

   Note: If the previously applied change created a custom role, then reverting the change does not delete the custom role. You can disable or delete the custom role if you no longer need it.

   To restore a recommendation that was dismissed, click Restore. The recommendation becomes visible on the IAM page in the Cloud Console. No roles or permissions are changed.

What's next

• Read the IAM recommender overview.
• Learn more about Recommender.
• Understand best practices for using the IAM recommender.
• Learn how to use IAM insights.
• Understand predefined roles and custom roles in IAM.