Attribute reference for IAM Conditions

This document describes supported attributes in a condition expression .

Supported condition attributes

The following sections summarize the supported attributes and indicate which Google Cloud services recognize each attribute.

Resource attributes

The following attributes relate to the resource that is the subject of the request.

For more details about resource attributes, see Resource attributes on this page.

Principal attributes

The following attributes relate to the principal making the request.

For more details about principal attributes, see Principal attributes on this page.

Request attributes

The following attributes relate to the details of the request.

For more details about request attributes, see Request attributes on this page.

Support for inherited conditions

Some types of Google Cloud resources don't allow conditions in their allow policies. However, you can add conditional role bindings at the organization, folder, or project level, and other resources will inherit those role bindings through the resource hierarchy . For details, see Resource types that accept conditional role bindings .

When you use attributes at the organization, folder, or project level, keep in mind that most attributes are available only for specific resource types. If part of a condition uses an attribute that is not available, then that part of the condition is never interpreted as granting access. For example, the condition resource.name.endsWith == devResource will never grant access to any IAM resource, because IAM resources don't provide the resource name.

To prevent this issue, use the resource type and resource service attributes described on this page to limit the scope of the condition. For example, the following condition evaluates to true for all resource types other than Compute Engine instances; in contrast, for Compute Engine instances, the condition checks resource name:

 resource.type != 'compute.googleapis.com/Disk' ||
    resource.name.endsWith('devResource') 

You don't need to limit the scope of conditions that check the tags attached to a resource . When a condition checks tag keys and values, it cannot check any other attributes, including the resource type and resource service.

Resource attributes

The resource service, resource type, and resource name attributes are typically used to change the scope of an access grant provided by the role binding. When a role contains permissions that apply to different resource-specific attributes, resource-based conditions can be used to grant a subset of the role's permissions for specific type(s) or for specific service(s).

resource.service attribute

The resource.service attribute lets you set a condition based on the Google Cloud service being used. For example, you could set a condition limiting a user's access to resources that use the cloudresourcemanager.googleapis.com service. For a list of supported values, see Resource service values .

You can use the resource.service attribute in allow policy role bindings.

resource.service == "compute.googleapis.com"

resource.type attribute

The resource.type attribute lets you set a condition based on the resource's type. For example, you could set a condition limiting a user's access to resources of the type storage.googleapis.com/Object . For a list of supported values, see Resource type values .

If your condition uses the resource.name attribute, we strongly recommend that you use the resource.type attribute to control which resource types the condition applies to. For details, see resource.name attribute on this page.

You can use the resource.type attribute in allow policy role bindings.

resource.type != "compute.googleapis.com/Image"

(resource.type == "compute.googleapis.com/Image" ||
 resource.type == "compute.googleapis.com/Disk")

resource.name attribute

The resource.name attribute lets you set a condition based on all or part of a resource name. For a list of resource name formats, see Resource name format .

The resource.name attribute is available only for specific resource types, which are listed in the table in this section. We strongly recommend that you limit the applicability of the condition to the intended resource type. If a role contains permissions for a resource type that does not provide the resource.name attribute, you should ensure that those permissions are not restricted by the part of the condition that checks resource.name .

The following example shows how to ensure this behavior. In this example, the condition allows access to all resource types except Cloud Storage buckets and objects. In contrast, for buckets and objects, the condition only allows access to the bucket example-bucket and the objects it contains:

 (resource.type != 'storage.googleapis.com/Bucket' &&
 resource.type != 'storage.googleapis.com/Object') ||
resource.name.startsWith('projects/_/buckets/example-bucket') 

Note that the first part of the condition checks whether the resource is neither a bucket nor an object. If the resource has a different type, then the entire condition evaluates to true , regardless of the resource name.

Also, note that the condition checks the resource.type attribute, not the resource.service attribute. There are a few benefits of checking the resource.type attribute:

• It limits the resource.name check to the appropriate set of resources. For example, if you want to grant access to Compute Engine instances with a specific name, it makes sense to exclude all resource types other than Compute Engine instances.
• It prevents the scope of the condition from changing if a service adds new resource types in the future.

Finally, note that the condition uses the startsWith() function to evaluate the resource name, rather than checking for equality with the == operator. Because the condition looks at the start of the resource name, it matches a bucket as well as the objects in that bucket. If it checked for equality, it would only match the bucket.

You cannot use wildcard characters such as * to match multiple resource names. Consider these alternatives:

• Use the extract() function to extract a value from a resource name. For example, you can extract a project ID from the resource name of a Compute Engine VM instance, then write a condition expression that refers to the project ID.

  For details, see Extracting values from attributes on this page.

• Use the startsWith() or endsWith() function to write a condition that evaluates the start or end of the resource name.

You can use the resource.name attribute in allow policy role bindings.

resource.name != "projects/_/buckets/secret-bucket-123"

resource.name.startsWith("projects/project-123/zones/us-east1-b/instances/prod-")

resource.name.startsWith("projects/_/buckets/my_bucket/objects/test-object-")

resource.name.endsWith(".jpg")

resource.name.extract("projects/{project}/")

Resource tags

The resource tag functions let you set a condition based on the tags that are attached to supported resources or inherited by those resources' descendants. For example, you can set a condition that grants a role only for resources that have the tag env: prod attached. To learn more about controlling access with tags, see Tags and access control .

Note:Conditions that check the tags for a resource and other attributes, such as the resource name or the timestamp of the request, are in preview. Such conditions are subject to the "Pre-GA Offerings Terms" in the General Service Terms section of the Service Specific Terms . For more information, see the launch stage descriptions .

Conditions that check the tags for a resource and don't check any other attributes are generally available.

Each tag consists of a key and a value. There are a few different types of identifiers for each key and value:

• A permanent ID , which is globally unique and can never be reused. For example, a tag key could have the permanent ID tagKeys/123456789012 , and a tag value could have the permanent ID tagValues/567890123456 .
• A short name . The short name for each key must be unique within the project or organization under which the key is defined, and the short name for each value must be unique for its associated key. For example, a tag key could have the short name env , and a tag value could have the short name prod .
• A namespaced name , which adds your organization's numeric ID or project's ID to the short name of a tag key. For example, a tag key created for an organization could have the namespaced name 123456789012/env . To learn how to get your organization ID, see Getting your organization resource ID . A tag key created for a project could have the namespaced name myproject/env . To learn how to get your project ID, see Identifying projects .

For guidance on choosing which type of identifier to use in your conditions, see Tag definitions and identifiers .

You can use tag-based conditions to conditionalize access to any resource. This includes resources with their own tags, as well as resources that inherit tags from other resources. To learn more about how tags are inherited through the resource hierarchy, see Tag inheritance .

However, certain areas of the Google Cloud console don't recognize allow policy role bindings with tag-based conditions. As a result, if you have a role with a tag-based condition, then the Google Cloud console might incorrectly prevent you from performing certain actions. If you encounter this issue, then use an alternate method, such as the gcloud CLI, to perform the action.

You can use tag-based conditions in the following:

• Allow policy role bindings
• Deny policy deny rules

You can use the following functions to set conditions based on tags:

Principal attributes

The principal attributes let you write conditions based on the principal that issued the request. With these attributes, you can refine the principals that a policy is enforced for.

You can use principal attributes in policy bindings for principal access boundary policies.

principal.type attribute

The principal.type attribute lets you set a condition based on the type of principal issuing the request. For example, you could add a condition to a policy binding for a principal access boundary policy to ensure that the policy is only enforced for service accounts.

You can use principal attributes in policy bindings for principal access boundary policies.

principal.type == "iam.googleapis.com/ServiceAccount"

principal.type in ["iam.googleapis.com/WorkspaceIdentity", "iam.googleapis.com/WorkforcePoolIdentity"]

principal.subject attribute

The principal.subject attribute lets you set a condition based on the principal issuing the request. For example, you could add a condition to a policy binding for a principal access boundary policy to ensure that the policy is only enforced for principals whose email addresses end with @example.com .

If you use the principal.subject attribute in a condition, we recommend also using the principal.type attribute to control which types of principals the condition applies to. This is because principal identifiers aren't necessarily unique across principal types. For example, the identifier example-user@example.com could identify a Google Account or a user in a workforce identity pool.

By using the principal.type attribute in addition to the principal.subject attribute, you can ensure that the condition only matches principals with the intended type. For example, the following expression matches Google Accounts whose email addresses end with @example.com :

 principal.type == 'iam.googleapis.com/WorkspaceIdentity' &&
principal.subject.endsWith('@example.com') 

You can use principal attributes in policy bindings for principal access boundary policies.

principal.subject.endsWith("@example.com")

principal.subject == "example-service-account@example-project.iam.gserviceaccount.com"

Request attributes

Request attributes enable you to create conditions that evaluate details about the request, such as its access level, its date and time, the destination IP address and port (for IAP TCP tunneling), or the expected URL path/host (for IAP and Cloud Run).

Access levels attribute

The access levels attribute enables users to set a condition requiring that a request meets one or more access levels in order to be authorized. You can use the access levels attribute in allow policy role bindings.

The access levels attribute is derived from attributes of the request, such as the origin IP address, device attributes, and the time of day. For example, an access level named fullyTrusted might require that the device making the request is owned by the company and has a screen lock. An onNetwork access level might require that the device making the request originates from a particular IP address range. See the Access Context Manager documentation for more information about access levels.

The access levels attribute is available only when you use Identity-Aware Proxy to access a tunnel instance, or to access a web application running on App Engine or Compute Engine backend services. More specifically, the access levels attribute is available only for requests that check one of these permissions:

• iap.tunnelInstances.accessViaIAP
• iap.webServiceVersions.accessViaIAP

You can use the access levels attribute when you conditionally grant the following predefined roles:

• IAP-secured Tunnel User ( roles/iap.tunnelResourceAccessor )

  Contains a single permission, iap.tunnelInstances.accessViaIAP .

• IAP-secured Web App User ( roles/iap.httpsResourceAccessor )

  Contains a single permission, iap.webServiceVersions.accessViaIAP .

You can also use the access levels attribute to conditionally grant a custom role that contains these permissions. The custom role must not contain any other permissions.

request.auth.access_levels attribute

 ACCESS_LEVEL_FULL_NAME 
in request.auth.access_levels

accessPolicies/ POLICY_NUMBER 
/accessLevels/ ACCESS_LEVEL 

"accessPolicies/199923665455/accessLevels/CorpNet"
    in request.auth.access_levels

API attributes

API attributes help you manage access based on data provided by a specific Google Cloud API or service. You can use API attributes in allow policy role bindings.

For example, when you use Cloud Storage to list the objects in a bucket , you can use the prefix parameter in the request to include only objects whose names begin with a specific prefix. If you use Credential Access Boundaries to downscope short-lived credentials, you can create a Credential Access Boundary that limits permissions to list objects by checking the API attribute storage.googleapis.com/objectListPrefix . This API attribute contains the value of the prefix parameter from the request.

For examples of when you might need to use API attributes in a condition, see the following pages:

• Limit permissions when listing objects
• Setting limits on granting roles

Not all services recognize API attributes. The following sections indicate which services recognize each API attribute.

Functions for API attributes

You can use the following function to work with API attributes:

Cloud Storage API attributes

Cloud Storage provides the following API attribute.

IAM API attributes

IAM provides the following API attribute:

Date/time attribute

The date/time attribute is used to set expirable, scheduled, or limited-duration access to Google Cloud resources. You can use date/time attributes in allow policy role bindings.

This attribute is supported for all Google Cloud services and resource types. To learn how to apply date/time conditions to resources that don't directly support them, see Support for inherited conditions on this page.

The request.time attribute contains the timestamp for the request. You can compare this timestamp to another timestamp, or to a duration of time.

The following sections list the functions that you can use to set conditions based on timestamps and durations.

Create, compare, and modify timestamps and durations

Extract information from a timestamp

The functions in this section let you extract information from a timestamp, such as the day of the week that the timestamp falls on.

In IAM Conditions, all timestamps are in UTC. However, you might want to extract information based on a different time zone. For example, you might want to know whether a UTC timestamp falls on a Monday in the time zone for Berlin, Germany.

To specify a different time zone, pass the time zone into the function. Use a name or UTC offset from the IETF Time Zone Database . For example, you could use Europe/Berlin or +01:00 for Central European Time (CET).

Destination IP/port attributes

The destination IP/port attribute enables users to manage access based on the internal destination IP address and port for a request. You can use destination IP/port attributes in allow policy role bindings.

For example, a Compute Engine VM instance might map the external IP address and port 132.168.42.21:3001 to the internal IP address and port 10.0.0.1:2300 for general usage. In contrast, the internal IP address and port 10.0.0.1:22 might only be available internally for administrative usage. You can use the destination IP/port attributes to grant different amounts of access based on the internal IP address and port.

For more information about TCP forwarding, see the Identity-Aware Proxy documentation .

destination.ip attribute

destination.ip == "10.0.0.1"

destination.ip != "10.0.0.1"

destination.port attribute

destination.port == 21

destination.port < 3001

Forwarding rule attributes

The forwarding rule attributes enable you to specify the types of forwarding rules that a principal can create. For example, you could allow a principal to create forwarding rules for internal Google Cloud load balancers , which handle traffic that originates inside a Google Cloud network, but not for external Google Cloud load balancers, which handle traffic that originates from the internet. You can use forwarding rule attributes in allow policy role bindings.

For Cloud Load Balancing, the forwarding rule attributes don't affect the ability to create other components of a Google Cloud load balancer, such as backend services, target proxies, health checks, and URL maps.

Supported functions

Supported resource types

This attribute is available for requests to create the following resource types:

¹ Uses the resource attributes for Compute Engine.

URL path/host attribute

The URL path/host attribute enables users to manage access based on the URL path and host of a request. For example, a condition could specify that https://example.com is the main application accessible by a general domain of users, while https://hr.example.com/admin is used to access a page in the application where only Human Resources admins can access this portion.

You can use the URL path/host attribute in allow policy role bindings.

request.path attribute

request.path == "/admin"

request.path == "/admin/payroll"

request.path.startsWith("/admin")

request.path.endsWith("/payroll.js")

request.host attribute

request.host == "www.example.com"

request.host == "hr.example.com"

request.host.endsWith("example.com")

Extract values from attributes

You can use the extract() function to extract a value from an attribute. For example, you can extract an arbitrary part of a resource name, then write a condition expression that refers to the text you extracted.

To use the extract() function, you provide an extraction template , which specifies the part of the attribute to extract. For example, if you want to extract a project ID from the resource name of a Compute Engine VM instance, you might use the template projects/{project}/ .

An extraction template contains the following parts:

• An identifier , enclosed in curly braces, that identifies the substring to extract.

  Choose a short, meaningful identifier that makes it clear what value you want to extract. You can use uppercase and lowercase letters from A to Z ; numeric digits; and underscores ( _ ).

  In the template projects/{project}/ , the identifier is project .

• Optional: A prefix , which must appear before the substring to extract.

  In the template projects/{project}/ , the prefix is projects/ .

• Optional: A suffix , which must appear after the substring to extract.

  In the template projects/{project}/ , the suffix is / .

The extract() function extracts different parts of the attribute based on whether the extraction template has a prefix, a suffix, or both:

If you specify a prefix or suffix that don't appear in the attribute, or if the suffix appears only before the prefix, the extract() function returns an empty string.

The following examples show the output from several different extraction templates. These examples refer to a resource name for a Cloud Storage object, projects/_/buckets/acme-orders-aaa/objects/data_lake/orders/ order_date=2019-11-03/aef87g87ae0876 :

If you extract a string that represents a date, you can use the date/time functions and operators on this page to convert the extracted value to a Timestamp . For examples, see Configuring resource-based access .