User Guide

Credentials and account types

Credentials are the means of identifying an application or user to a service or API. Credentials can be obtained with three different types of accounts: service accounts, user accounts and external accounts.

Credentials from service accounts identify a particular application. These types of credentials are used in server-to-server use cases, such as accessing a database. This library primarily focuses on service account credentials.

Credentials from user accounts are obtained by asking the user to authorize access to their data. These types of credentials are used in cases where your application needs access to a user’s data in another service, such as accessing a user’s documents in Google Drive. This library provides no support for obtaining user credentials, but does provide limited support for using user credentials.

Credentials from external accounts (workload identity federation) are used to identify a particular application from an on-prem or non-Google Cloud platform including Amazon Web Services (AWS), Microsoft Azure or any identity provider that supports OpenID Connect (OIDC).

Obtaining credentials

Application default credentials

Google Application Default Credentials abstracts authentication across the different Google Cloud Platform hosting environments. When running on any Google Cloud hosting environment or when running locally with the Google Cloud SDK installed, default() can automatically determine the credentials from the environment:

import google.auth

credentials, project = google.auth.default()

If your application requires specific scopes:

credentials, project = google.auth.default(
    scopes=['https://www.googleapis.com/auth/cloud-platform'])

Application Default Credentials also support workload identity federation to access Google Cloud resources from non-Google Cloud platforms including Amazon Web Services (AWS), Microsoft Azure or any identity provider that supports OpenID Connect (OIDC). Workload identity federation is recommended for non-Google Cloud environments as it avoids the need to download, manage and store service account private keys locally.

Service account private key files

A service account private key file can be used to obtain credentials for a service account. You can create a private key using the Credentials page of the Google Cloud Console. Once you have a private key you can either obtain credentials one of three ways:

  1. Set the GOOGLE_APPLICATION_CREDENTIALS environment variable to the full path to your service account private key file

    $ export GOOGLE_APPLICATION_CREDENTIALS=/path/to/key.json
    

    Then, use application default credentials. default() checks for the GOOGLE_APPLICATION_CREDENTIALS environment variable before all other checks, so this will always use the credentials you explicitly specify.

  2. Use service_account.Credentials.from_service_account_file:

    from google.oauth2 import service_account
    
    credentials = service_account.Credentials.from_service_account_file(
        '/path/to/key.json')
    
    scoped_credentials = credentials.with_scopes(
        ['https://www.googleapis.com/auth/cloud-platform'])
    
  3. Use service_account.Credentials.from_service_account_info:

    import json
    
    from google.oauth2 import service_account
    
    json_acct_info = json.loads(function_to_get_json_creds())
    credentials = service_account.Credentials.from_service_account_info(
        json_acct_info)
    
    scoped_credentials = credentials.with_scopes(
        ['https://www.googleapis.com/auth/cloud-platform'])
    

Warning

Private keys must be kept secret. If you expose your private key it is recommended to revoke it immediately from the Google Cloud Console.

Compute Engine, Container Engine, and the App Engine flexible environment

Applications running on Compute Engine, Container Engine, or the App Engine flexible environment can obtain credentials provided by Compute Engine service accounts. When running on these platforms you can obtain credentials for the service account one of two ways:

  1. Use application default credentials. default() will automatically detect if these credentials are available.

  2. Use compute_engine.Credentials:

    from google.auth import compute_engine
    
    credentials = compute_engine.Credentials()
    

The App Engine standard environment

Applications running on the App Engine standard environment can obtain credentials provided by the App Engine App Identity API. You can obtain credentials one of two ways:

  1. Use application default credentials. default() will automatically detect if these credentials are available.

  2. Use app_engine.Credentials:

    from google.auth import app_engine
    
    credentials = app_engine.Credentials()
    

In order to make authenticated requests in the App Engine environment using the credentials and transports provided by this library, you need to follow a few additional steps:

  1. If you are using the google.auth.transport.requests transport, vendor in the requests-toolbelt library into your app, and enable the App Engine monkeypatch. Refer App Engine documentation for more details on this.

  2. To make HTTPS calls, enable the ssl library for your app by adding the following configuration to the app.yaml file:

    libraries:
    - name: ssl
      version: latest
    
  3. Enable billing for your App Engine project. Then enable socket support for your app. This can be achieved by setting an environment variable in the app.yaml file:

    env_variables:
      GAE_USE_SOCKETS_HTTPLIB : 'true'
    

User credentials

User credentials are typically obtained via OAuth 2.0. This library does not provide any direct support for obtaining user credentials, however, you can use user credentials with this library. You can use libraries such as oauthlib to obtain the access token. After you have an access token, you can create a google.oauth2.credentials.Credentials instance:

import google.oauth2.credentials

credentials = google.oauth2.credentials.Credentials(
    'access_token')

If you obtain a refresh token, you can also specify the refresh token and token URI to allow the credentials to be automatically refreshed:

credentials = google.oauth2.credentials.Credentials(
    'access_token',
    refresh_token='refresh_token',
    token_uri='token_uri',
    client_id='client_id',
    client_secret='client_secret')

There is a separate library, google-auth-oauthlib, that has some helpers for integrating with requests-oauthlib to provide support for obtaining user credentials. You can use google_auth_oauthlib.helpers.credentials_from_session() to obtain google.oauth2.credentials.Credentials from a requests_oauthlib.OAuth2Session as above:

from google_auth_oauthlib.helpers import credentials_from_session

google_auth_credentials = credentials_from_session(oauth2session)

You can also use google_auth_oauthlib.flow.Flow to perform the OAuth 2.0 Authorization Grant Flow to obtain credentials using requests-oauthlib.

External credentials (Workload identity federation)

Using workload identity federation, your application can access Google Cloud resources from Amazon Web Services (AWS), Microsoft Azure or any identity provider that supports OpenID Connect (OIDC).

Traditionally, applications running outside Google Cloud have used service account keys to access Google Cloud resources. Using identity federation, you can allow your workload to impersonate a service account. This lets you access Google Cloud resources directly, eliminating the maintenance and security burden associated with service account keys.

Accessing resources from AWS

In order to access Google Cloud resources from Amazon Web Services (AWS), the following requirements are needed:

  • A workload identity pool needs to be created.

  • AWS needs to be added as an identity provider in the workload identity pool (The Google organization policy needs to allow federation from AWS).

  • Permission to impersonate a service account needs to be granted to the external identity.

  • A credential configuration file needs to be generated. Unlike service account credential files, the generated credential configuration file will only contain non-sensitive metadata to instruct the library on how to retrieve external subject tokens and exchange them for service account access tokens.

  • If you want to use IDMSv2, then below field needs to be added to credential_source section of credential configuration. “imdsv2_session_token_url”: “http://169.254.169.254/latest/api/token

Follow the detailed instructions on how to Configure Workload Identity Federation from AWS.

Accessing resources from Microsoft Azure

In order to access Google Cloud resources from Microsoft Azure, the following requirements are needed:

  • A workload identity pool needs to be created.

  • Azure needs to be added as an identity provider in the workload identity pool (The Google organization policy needs to allow federation from Azure).

  • The Azure tenant needs to be configured for identity federation.

  • Permission to impersonate a service account needs to be granted to the external identity.

  • A credential configuration file needs to be generated. Unlike service account credential files, the generated credential configuration file will only contain non-sensitive metadata to instruct the library on how to retrieve external subject tokens and exchange them for service account access tokens.

Follow the detailed instructions on how to Configure Workload Identity Federation from Microsoft Azure.

Accessing resources from an OIDC identity provider

In order to access Google Cloud resources from an identity provider that supports OpenID Connect (OIDC), the following requirements are needed:

  • A workload identity pool needs to be created.

  • An OIDC identity provider needs to be added in the workload identity pool (The Google organization policy needs to allow federation from the identity provider).

  • Permission to impersonate a service account needs to be granted to the external identity.

  • A credential configuration file needs to be generated. Unlike service account credential files, the generated credential configuration file will only contain non-sensitive metadata to instruct the library on how to retrieve external subject tokens and exchange them for service account access tokens.

For OIDC providers, the Auth library can retrieve OIDC tokens either from a local file location (file-sourced credentials) or from a local server (URL-sourced credentials).

  • For file-sourced credentials, a background process needs to be continuously refreshing the file location with a new OIDC token prior to expiration. For tokens with one hour lifetimes, the token needs to be updated in the file every hour. The token can be stored directly as plain text or in JSON format.

  • For URL-sourced credentials, a local server needs to host a GET endpoint to return the OIDC token. The response can be in plain text or JSON. Additional required request headers can also be specified.

Follow the detailed instructions on how to Configure Workload Identity Federation from an OIDC identity provider.

Using Executable-sourced credentials with OIDC and SAML

Executable-sourced credentials For executable-sourced credentials, a local executable is used to retrieve the 3rd party token. The executable must handle providing a valid, unexpired OIDC ID token or SAML assertion in JSON format to stdout.

To use executable-sourced credentials, the GOOGLE_EXTERNAL_ACCOUNT_ALLOW_EXECUTABLES environment variable must be set to 1.

To generate an executable-sourced workload identity configuration, run the following command:

# Generate a configuration file for executable-sourced credentials.
gcloud iam workload-identity-pools create-cred-config \
    projects/$PROJECT_NUMBER/locations/global/workloadIdentityPools/$POOL_ID/providers/$PROVIDER_ID \
    --service-account=$SERVICE_ACCOUNT_EMAIL \
    --subject-token-type=$SUBJECT_TOKEN_TYPE \
    # The absolute path for the program, including arguments.
    # e.g. --executable-command="/path/to/command --foo=bar"
    --executable-command=$EXECUTABLE_COMMAND \
    # Optional argument for the executable timeout. Defaults to 30s.
    # --executable-timeout-millis=$EXECUTABLE_TIMEOUT \
    # Optional argument for the absolute path to the executable output file.
    # See below on how this argument impacts the library behaviour.
    # --executable-output-file=$EXECUTABLE_OUTPUT_FILE \
    --output-file /path/to/generated/config.json

Where the following variables need to be substituted: - $PROJECT_NUMBER: The Google Cloud project number. - $POOL_ID: The workload identity pool ID. - $PROVIDER_ID: The OIDC or SAML provider ID. - $SERVICE_ACCOUNT_EMAIL: The email of the service account to impersonate. - $SUBJECT_TOKEN_TYPE: The subject token type. - $EXECUTABLE_COMMAND: The full command to run, including arguments. Must be an absolute path to the program.

The --executable-timeout-millis flag is optional. This is the duration for which the auth library will wait for the executable to finish, in milliseconds. Defaults to 30 seconds when not provided. The maximum allowed value is 2 minutes. The minimum is 5 seconds.

The --executable-output-file flag is optional. If provided, the file path must point to the 3PI credential response generated by the executable. This is useful for caching the credentials. By specifying this path, the Auth libraries will first check for its existence before running the executable. By caching the executable JSON response to this file, it improves performance as it avoids the need to run the executable until the cached credentials in the output file are expired. The executable must handle writing to this file - the auth libraries will only attempt to read from this location. The format of contents in the file should match the JSON format expected by the executable shown below.

To retrieve the 3rd party token, the library will call the executable using the command specified. The executable’s output must adhere to the response format specified below. It must output the response to stdout.

A sample successful executable OIDC response:

{
  "version": 1,
  "success": true,
  "token_type": "urn:ietf:params:oauth:token-type:id_token",
  "id_token": "HEADER.PAYLOAD.SIGNATURE",
  "expiration_time": 1620499962
}

A sample successful executable SAML response:

{
  "version": 1,
  "success": true,
  "token_type": "urn:ietf:params:oauth:token-type:saml2",
  "saml_response": "...",
  "expiration_time": 1620499962
}

A sample executable error response:

{
  "version": 1,
  "success": false,
  "code": "401",
  "message": "Caller not authorized."
}

These are all required fields for an error response. The code and message fields will be used by the library as part of the thrown exception.

Response format fields summary:

  • version: The version of the JSON output. Currently only version 1 is supported.

  • success: The status of the response.
    • When true, the response must contain the 3rd party token, token type, and expiration. The executable must also exit with exit code 0.

    • When false, the response must contain the error code and message fields and exit with a non-zero value.

  • token_type: The 3rd party subject token type. Must be
    • urn:ietf:params:oauth:token-type:jwt

    • urn:ietf:params:oauth:token-type:id_token

    • urn:ietf:params:oauth:token-type:saml2

  • id_token: The 3rd party OIDC token.

  • saml_response: The 3rd party SAML response.

  • expiration_time: The 3rd party subject token expiration time in seconds (unix epoch time).

  • code: The error code string.

  • message: The error message.

All response types must include both the version and success fields. Successful responses must include the token_type, and one of id_token or saml_response. expiration_time is optional. If the output file does not contain the expiration_time field, the response will be considered expired and the executable will be called. Error responses must include both the code and message fields.

The library will populate the following environment variables when the executable is run: GOOGLE_EXTERNAL_ACCOUNT_AUDIENCE: The audience field from the credential configuration. Always present. GOOGLE_EXTERNAL_ACCOUNT_IMPERSONATED_EMAIL: The service account email. Only present when service account impersonation is used. GOOGLE_EXTERNAL_ACCOUNT_OUTPUT_FILE: The output file location from the credential configuration. Only present when specified in the credential configuration.

These environment variables can be used by the executable to avoid hard-coding these values.

Security considerations

The following security practices are highly recommended: Access to the script should be restricted as it will be displaying credentials to stdout. This ensures that rogue processes do not gain access to the script. The configuration file should not be modifiable. Write access should be restricted to avoid processes modifying the executable command portion.

Given the complexity of using executable-sourced credentials, it is recommended to use the existing supported mechanisms (file-sourced/URL-sourced) for providing 3rd party credentials unless they do not meet your specific requirements.

You can now use the Auth library to call Google Cloud resources from an OIDC or SAML provider.

Accessing resources using a custom supplier with OIDC or SAML

This library also allows for a custom implementation of google.auth.identity_pool.SubjectTokenSupplier to be specificed when creating a google.auth.identity_pool.Credential. The supplier must return a valid OIDC or SAML2.0 subject token, which will then be exchanged for a Google Cloud access token. If an error occurs during token retrieval, the supplier should return a google.auth.exceptions.RefreshError and indicate via the error whether the subject token retrieval is retryable. Any call to the supplier from the Identity Pool credential will send a google.auth.external_account.SupplierContext object, which contains the requested audience and subject type. Additionally, the credential will send the google.auth.transport.requests.Request passed in the credential refresh call which can be used to make HTTP requests.:

from google.auth import exceptions
from google.auth import identity_pool

class CustomSubjectTokenSupplier(identity_pool.SubjectTokenSupplier):

    def get_subject_token(self, context, request):
        audience = context.audience
        subject_token_type = context.subject_token_type
        try:
            # Attempt to return the valid subject token of the requested type for the requested audience.
        except Exception as e:
            # If token retrieval fails, raise a refresh error, setting retryable to true if the client should
            # attempt to retrieve the subject token again.
            raise exceptions.RefreshError(e, retryable=True)

supplier = CustomSubjectTokenSupplier()

credentials = identity_pool.Credentials(
    AUDIENCE, # Set GCP Audience.
    "urn:ietf:params:aws:token-type:jwt", # Set subject token type.
    subject_token_supplier=supplier, # Set supplier.
    scopes=SCOPES # Set desired scopes.
)

Where the audience is: ///iam.googleapis.com/projects/PROJECT_NUMBER/locations/global/workloadIdentityPools/POOL_ID/providers/PROVIDER_ID Where the following variables need to be substituted:

  • $PROJECT_NUMBER: The project number.

  • $POOL_ID: The workload pool ID.

  • $PROVIDER_ID: The provider ID.

The values for audience, service account impersonation URL, and any other builder field can also be found by generating a credential configuration file with the gcloud CLI.

Accessing resources using a custom supplier with AWS

This library also allows for a custom implementation of google.auth.aws.AwsSecurityCredentialsSupplier to be specificed when creating a google.auth.aws.Credential. The supplier must return valid AWS security credentials, which will then be exchanged for a Google Cloud access token. If an error occurs during credential retrieval, the supplier should return a google.auth.exceptions.RefreshError and indicate via the error whether the credential retrieval is retryable. Any call to the supplier from the Identity Pool credential will send a google.auth.external_account.SupplierContext object, which contains the requested audience and subject type. Additionally, the credential will send the google.auth.transport.requests.Request passed in the credential refresh call which can be used to make HTTP requests.:

from google.auth import aws
from google.auth import exceptions

class CustomAwsSecurityCredentialsSupplier(aws.AwsSecurityCredentialsSupplier):

    def get_aws_security_credentials(self, context, request):
        audience = context.audience
        try:
            # Return valid AWS security credentials. These credentials are not cached by
            # the google credential, so caching should be implemented in the supplier.
            return aws.AwsSecurityCredentials(ACCESS_KEY_ID, SECRET_ACCESS_KEY, SESSION_TOKEN)
        except Exception as e:
            # If credentials retrieval fails, raise a refresh error, setting retryable to true if the client should
            # attempt to retrieve the subject token again.
            raise exceptions.RefreshError(e, retryable=True)

    def get_aws_region(self, context, request):
        # Return active AWS region.

supplier = CustomAwsSecurityCredentialsSupplier()

credentials = aws.Credentials(
    AUDIENCE, # Set GCP Audience.
    "urn:ietf:params:aws:token-type:aws4_request", # Set AWS subject token type.
    aws_security_token_supplier=supplier, # Set supplier.
    scopes=SCOPES # Set desired scopes.
)

Where the audience is: ///iam.googleapis.com/projects/PROJECT_NUMBER/locations/global/workloadIdentityPools/POOL_ID/providers/PROVIDER_ID Where the following variables need to be substituted:

  • $PROJECT_NUMBER: The project number.

  • $POOL_ID: The workload pool ID.

  • $PROVIDER_ID: The provider ID.

The values for audience, service account impersonation URL, and any other builder field can also be found by generating a credential configuration file with the gcloud CLI.

Using External Identities

External identities (AWS, Azure and OIDC identity providers) can be used with Application Default Credentials. In order to use external identities with Application Default Credentials, you need to generate the JSON credentials configuration file for your external identity. Once generated, store the path to this file in the GOOGLE_APPLICATION_CREDENTIALS environment variable.

$ export GOOGLE_APPLICATION_CREDENTIALS=/path/to/config.json

The library can now automatically choose the right type of client and initialize credentials from the context provided in the configuration file:

import google.auth

credentials, project = google.auth.default()

When using external identities with Application Default Credentials, the roles/browser role needs to be granted to the service account. The Cloud Resource Manager API should also be enabled on the project. This is needed since default() will try to auto-discover the project ID from the current environment using the impersonated credential. Otherwise, the project ID will resolve to None. You can override the project detection by setting the GOOGLE_CLOUD_PROJECT environment variable.

You can also explicitly initialize external account clients using the generated configuration file.

For Azure and OIDC providers, use identity_pool.Credentials.from_info or identity_pool.Credentials.from_file:

import json

from google.auth import identity_pool

json_config_info = json.loads(function_to_get_json_config())
credentials = identity_pool.Credentials.from_info(json_config_info)
scoped_credentials = credentials.with_scopes(
    ['https://www.googleapis.com/auth/cloud-platform'])

For AWS providers, use aws.Credentials.from_info or aws.Credentials.from_file:

import json

from google.auth import aws

json_config_info = json.loads(function_to_get_json_config())
credentials = aws.Credentials.from_info(json_config_info)
scoped_credentials = credentials.with_scopes(
    ['https://www.googleapis.com/auth/cloud-platform'])

Security considerations

Note that this library does not perform any validation on the token_url, token_info_url, or service_account_impersonation_url fields of the credential configuration. It is not recommended to use a credential configuration that you did not generate with the gcloud CLI unless you verify that the URL fields point to a googleapis.com domain.

External credentials (Workforce identity federation)

Workforce identity federation lets you use an external identify provider (IdP) to authenticate and authorize a workforce—a group of users, such as employees, partners, and contractors—using IAM, so that the users can access Google Cloud services. Workforce identity federation extends Google Cloud’s identity capabilities to support syncless, attribute-based single sign on.

With workforce identity federation, your workforce can access Google Cloud resources using an external identity provider (IdP) that supports OpenID Connect (OIDC) or SAML 2.0 such as Azure Active Directory (Azure AD), Active Directory Federation Services (AD FS), Okta, and others.

Accessing resources using an OIDC or SAML 2.0 identity provider

In order to access Google Cloud resources from an identity provider that supports OpenID Connect (OIDC), the following requirements are needed:

  • A workforce identity pool needs to be created.

  • An OIDC or SAML 2.0 identity provider needs to be added in the workforce pool.

Follow the detailed instructions on how to configure workforce identity federation.

After configuring an OIDC or SAML 2.0 provider, a credential configuration file needs to be generated. The generated credential configuration file contains non-sensitive metadata to instruct the library on how to retrieve external subject tokens and exchange them for GCP access tokens. The configuration file can be generated by using the gcloud CLI.

The Auth library can retrieve external subject tokens from a local file location (file-sourced credentials), from a local server (URL-sourced credentials) or by calling an executable (executable-sourced credentials).

File-sourced credentials

For file-sourced credentials, a background process needs to be continuously refreshing the file location with a new subject token prior to expiration. For tokens with one hour lifetimes, the token needs to be updated in the file every hour. The token can be stored directly as plain text or in JSON format.

To generate a file-sourced OIDC configuration, run the following command:

# Generate an OIDC configuration file for file-sourced credentials.
gcloud iam workforce-pools create-cred-config \
    locations/global/workforcePools/$WORKFORCE_POOL_ID/providers/$PROVIDER_ID \
    --subject-token-type=urn:ietf:params:oauth:token-type:id_token \
    --credential-source-file=$PATH_TO_OIDC_ID_TOKEN \
    --workforce-pool-user-project=$WORKFORCE_POOL_USER_PROJECT \
    # Optional arguments for file types. Default is "text":
    # --credential-source-type "json" \
    # Optional argument for the field that contains the OIDC credential.
    # This is required for json.
    # --credential-source-field-name "id_token" \
    --output-file=/path/to/generated/config.json

Where the following variables need to be substituted:

  • $WORKFORCE_POOL_ID: The workforce pool ID.

  • $PROVIDER_ID: The provider ID.

  • $PATH_TO_OIDC_ID_TOKEN: The file path used to retrieve the OIDC token.

  • $WORKFORCE_POOL_USER_PROJECT: The project number associated with the workforce pools user project.

To generate a file-sourced SAML configuration, run the following command:

# Generate a SAML configuration file for file-sourced credentials.
gcloud iam workforce-pools create-cred-config \
    locations/global/workforcePools/$WORKFORCE_POOL_ID/providers/$PROVIDER_ID \
    --credential-source-file=$PATH_TO_SAML_ASSERTION \
    --subject-token-type=urn:ietf:params:oauth:token-type:saml2 \
    --workforce-pool-user-project=$WORKFORCE_POOL_USER_PROJECT \
    --output-file=/path/to/generated/config.json

Where the following variables need to be substituted:

  • $WORKFORCE_POOL_ID: The workforce pool ID.

  • $PROVIDER_ID: The provider ID.

  • $PATH_TO_SAML_ASSERTION: The file path used to retrieve the base64-encoded SAML assertion.

  • $WORKFORCE_POOL_USER_PROJECT: The project number associated with the workforce pools user project.

These commands generate the configuration file in the specified output file.

URL-sourced credentials

For URL-sourced credentials, a local server needs to host a GET endpoint to return the OIDC token. The response can be in plain text or JSON. Additional required request headers can also be specified.

To generate a URL-sourced OIDC workforce identity configuration, run the following command:

# Generate an OIDC configuration file for URL-sourced credentials.
gcloud iam workforce-pools create-cred-config \
    locations/global/workforcePools/$WORKFORCE_POOL_ID/providers/$PROVIDER_ID \
    --subject-token-type=urn:ietf:params:oauth:token-type:id_token \
    --credential-source-url=$URL_TO_RETURN_OIDC_ID_TOKEN \
    --credential-source-headers $HEADER_KEY=$HEADER_VALUE \
    --workforce-pool-user-project=$WORKFORCE_POOL_USER_PROJECT \
    --output-file=/path/to/generated/config.json

Where the following variables need to be substituted:

  • $WORKFORCE_POOL_ID: The workforce pool ID.

  • $PROVIDER_ID: The provider ID.

  • $URL_TO_RETURN_OIDC_ID_TOKEN: The URL of the local server endpoint.

  • $HEADER_KEY and $HEADER_VALUE: The additional header key/value pairs to pass along the GET request to $URL_TO_GET_OIDC_TOKEN, e.g. Metadata-Flavor=Google.

  • $WORKFORCE_POOL_USER_PROJECT: The project number associated with the workforce pools user project.

To generate a URL-sourced SAML configuration, run the following command:

# Generate a SAML configuration file for file-sourced credentials.
gcloud iam workforce-pools create-cred-config \
    locations/global/workforcePools/$WORKFORCE_POOL_ID/providers/$PROVIDER_ID \
    --subject-token-type=urn:ietf:params:oauth:token-type:saml2 \
    --credential-source-url=$URL_TO_GET_SAML_ASSERTION \
    --credential-source-headers $HEADER_KEY=$HEADER_VALUE \
    --workforce-pool-user-project=$WORKFORCE_POOL_USER_PROJECT \
    --output-file=/path/to/generated/config.json

These commands generate the configuration file in the specified output file.

Where the following variables need to be substituted:

  • $WORKFORCE_POOL_ID: The workforce pool ID.

  • $PROVIDER_ID: The provider ID.

  • $URL_TO_GET_SAML_ASSERTION: The URL of the local server endpoint.

  • $HEADER_KEY and $HEADER_VALUE: The additional header key/value pairs to pass along the GET request to $URL_TO_GET_SAML_ASSERTION, e.g. Metadata-Flavor=Google.

  • $WORKFORCE_POOL_USER_PROJECT: The project number associated with the workforce pools user project.

Using Executable-sourced workforce credentials with OIDC and SAML

Executable-sourced credentials For executable-sourced credentials, a local executable is used to retrieve the 3rd party token. The executable must handle providing a valid, unexpired OIDC ID token or SAML assertion in JSON format to stdout.

To use executable-sourced credentials, the GOOGLE_EXTERNAL_ACCOUNT_ALLOW_EXECUTABLES environment variable must be set to 1.

To generate an executable-sourced workforce identity configuration, run the following command:

# Generate a configuration file for executable-sourced credentials.
gcloud iam workforce-pools create-cred-config \
    locations/global/workforcePools/$WORKFORCE_POOL_ID/providers/$PROVIDER_ID \
    --subject-token-type=$SUBJECT_TOKEN_TYPE \
    # The absolute path for the program, including arguments.
    # e.g. --executable-command="/path/to/command --foo=bar"
    --executable-command=$EXECUTABLE_COMMAND \
    # Optional argument for the executable timeout. Defaults to 30s.
    # --executable-timeout-millis=$EXECUTABLE_TIMEOUT \
    # Optional argument for the absolute path to the executable output file.
    # See below on how this argument impacts the library behaviour.
    # --executable-output-file=$EXECUTABLE_OUTPUT_FILE \
    --workforce-pool-user-project=$WORKFORCE_POOL_USER_PROJECT \
    --output-file /path/to/generated/config.json

Where the following variables need to be substituted:

  • $WORKFORCE_POOL_ID: The workforce pool ID.

  • $PROVIDER_ID: The provider ID.

  • $SUBJECT_TOKEN_TYPE: The subject token type.

  • $EXECUTABLE_COMMAND: The full command to run, including arguments. Must be an absolute path to the program.

  • $WORKFORCE_POOL_USER_PROJECT: The project number associated with the workforce pools user project.

The --executable-timeout-millis flag is optional. This is the duration for which the auth library will wait for the executable to finish, in milliseconds. Defaults to 30 seconds when not provided. The maximum allowed value is 2 minutes. The minimum is 5 seconds.

The --executable-output-file flag is optional. If provided, the file path must point to the 3rd party credential response generated by the executable. This is useful for caching the credentials. By specifying this path, the Auth libraries will first check for its existence before running the executable. By caching the executable JSON response to this file, it improves performance as it avoids the need to run the executable until the cached credentials in the output file are expired. The executable must handle writing to this file - the auth libraries will only attempt to read from this location. The format of contents in the file should match the JSON format expected by the executable shown below.

To retrieve the 3rd party token, the library will call the executable using the command specified. The executable’s output must adhere to the response format specified below. It must output the response to stdout.

Refer to the using executable-sourced credentials with Workload Identity Federation above for the executable response specification.

Accessing resources using a custom supplier with OIDC or SAML

This library also allows for a custom implementation of google.auth.identity_pool.SubjectTokenSupplier to be specificed when creating a google.auth.identity_pool.Credential. The supplier must return a valid OIDC or SAML2.0 subject token, which will then be exchanged for a Google Cloud access token. If an error occurs during token retrieval, the supplier should return a google.auth.exceptions.RefreshError and indicate via the error whether the subject token retrieval is retryable. Any call to the supplier from the Identity Pool credential will send a google.auth.external_account.SupplierContext object, which contains the requested audience and subject type. Additionally, the credential will send the google.auth.transport.requests.Request passed in the credential refresh call which can be used to make HTTP requests.:

from google.auth import exceptions
from google.auth import identity_pool

class CustomSubjectTokenSupplier(identity_pool.SubjectTokenSupplier):

    def get_subject_token(self, context, request):
        audience = context.audience
        subject_token_type = context.subject_token_type
        try:
            # Attempt to return the valid subject token of the requested type for the requested audience.
        except Exception as e:
            # If token retrieval fails, raise a refresh error, setting retryable to true if the client should
            # attempt to retrieve the subject token again.
            raise exceptions.RefreshError(e, retryable=True)


supplier = CustomSubjectTokenSupplier()

credentials = identity_pool.Credentials(
    AUDIENCE, # Set GCP Audience.
    "urn:ietf:params:aws:token-type:jwt", # Set subject token type.
    subject_token_supplier=supplier, # Set supplier.
    scopes=SCOPES, # Set desired scopes.
    workforce_pool_user_project=USER_PROJECT # Set workforce pool user project.
)

Where the audience is: //iam.googleapis.com/locations/global/workforcePools/$WORKFORCE_POOL_ID/providers/$PROVIDER_ID Where the following variables need to be substituted:

  • $WORKFORCE_POOL_ID: The workforce pool ID.

  • $PROVIDER_ID: The provider ID.

and the workforce pool user project is the project number associated with the workforce pools user project.

The values for audience, service account impersonation URL, and any other builder field can also be found by generating a credential configuration file with the gcloud CLI.

Security considerations

The following security practices are highly recommended:

  • Access to the script should be restricted as it will be displaying credentials to stdout. This ensures that rogue processes do not gain access to the script.

  • The configuration file should not be modifiable. Write access should be restricted to avoid processes modifying the executable command portion.

Given the complexity of using executable-sourced credentials, it is recommended to use the existing supported mechanisms (file-sourced/URL-sourced) for providing 3rd party credentials unless they do not meet your specific requirements.

You can now use the Auth library to call Google Cloud resources from an OIDC or SAML provider.

Note that this library does not perform any validation on the token_url, token_info_url, or service_account_impersonation_url fields of the credential configuration. It is not recommended to use a credential configuration that you did not generate with the gcloud CLI unless you verify that the URL fields point to a googleapis.com domain.

Impersonated credentials

Impersonated Credentials allows one set of credentials issued to a user or service account to impersonate another. The source credentials must be granted the “Service Account Token Creator” IAM role.

from google.auth import impersonated_credentials

target_scopes = ['https://www.googleapis.com/auth/devstorage.read_only']
source_credentials = service_account.Credentials.from_service_account_file(
    '/path/to/svc_account.json',
    scopes=target_scopes)

target_credentials = impersonated_credentials.Credentials(
    source_credentials=source_credentials,
    target_principal='impersonated-account@_project_.iam.gserviceaccount.com',
    target_scopes=target_scopes,
    lifetime=500)
client = storage.Client(credentials=target_credentials)
buckets = client.list_buckets(project='your_project')
for bucket in buckets:
    print(bucket.name)

In the example above source_credentials does not have direct access to list buckets in the target project. Using ImpersonatedCredentials will allow the source_credentials to assume the identity of a target_principal that does have access.

Downscoped credentials

Downscoping with Credential Access Boundaries is used to restrict the Identity and Access Management (IAM) permissions that a short-lived credential can use.

To downscope permissions of a source credential, a Credential Access Boundary that specifies which resources the new credential can access, as well as an upper bound on the permissions that are available on each resource, has to be defined. A downscoped credential can then be instantiated using the source_credential and the Credential Access Boundary.

The common pattern of usage is to have a token broker with elevated access generate these downscoped credentials from higher access source credentials and pass the downscoped short-lived access tokens to a token consumer via some secure authenticated channel for limited access to Google Cloud Storage resources.

Token broker

import google.auth

from google.auth import downscoped
from google.auth.transport import requests

# Initialize the credential access boundary rules.
available_resource = '//storage.googleapis.com/projects/_/buckets/bucket-123'
available_permissions = ['inRole:roles/storage.objectViewer']
availability_expression = (
    "resource.name.startsWith('projects/_/buckets/bucket-123/objects/customer-a')"
)

availability_condition = downscoped.AvailabilityCondition(
    availability_expression)
rule = downscoped.AccessBoundaryRule(
    available_resource=available_resource,
    available_permissions=available_permissions,
    availability_condition=availability_condition)
credential_access_boundary = downscoped.CredentialAccessBoundary(
    rules=[rule])

# Retrieve the source credentials via ADC.
source_credentials, _ = google.auth.default()

# Create the downscoped credentials.
downscoped_credentials = downscoped.Credentials(
    source_credentials=source_credentials,
    credential_access_boundary=credential_access_boundary)

# Refresh the tokens.
downscoped_credentials.refresh(requests.Request())

# These values will need to be passed to the Token Consumer.
access_token = downscoped_credentials.token
expiry = downscoped_credentials.expiry

For example, a token broker can be set up on a server in a private network. Various workloads (token consumers) in the same network will send authenticated requests to that broker for downscoped tokens to access or modify specific google cloud storage buckets.

The broker will instantiate downscoped credentials instances that can be used to generate short lived downscoped access tokens that can be passed to the token consumer. These downscoped access tokens can be injected by the consumer into google.oauth2.Credentials and used to initialize a storage client instance to access Google Cloud Storage resources with restricted access.

Token Consumer

import google.oauth2

from google.auth.transport import requests
from google.cloud import storage

# Downscoped token retrieved from token broker.
# The `get_token_from_broker` callable requests a token and an expiry
# from the token broker.
downscoped_token, expiry = get_token_from_broker(
    requests.Request(),
    scopes=['https://www.googleapis.com/auth/cloud-platform'])

# Create the OAuth credentials from the downscoped token and pass a
# refresh handler to handle token expiration. Passing the original
# downscoped token or the expiry here is optional, as the refresh_handler
# will generate the downscoped token on demand.
credentials = google.oauth2.credentials.Credentials(
    downscoped_token,
    expiry=expiry,
    scopes=['https://www.googleapis.com/auth/cloud-platform'],
    refresh_handler=get_token_from_broker)

# Initialize a storage client with the oauth2 credentials.
storage_client = storage.Client(
    project='my_project_id', credentials=credentials)
# Call GCS APIs.
# The token broker has readonly access to objects starting with "customer-a"
# in bucket "bucket-123".
bucket = storage_client.bucket('bucket-123')
blob = bucket.blob('customer-a-data.txt')
print(blob.download_as_bytes().decode("utf-8"))

Another reason to use downscoped credentials is to ensure tokens in flight always have the least privileges, e.g. Principle of Least Privilege.

# Create the downscoped credentials.
downscoped_credentials = downscoped.Credentials(
    # source_credentials have elevated access but only a subset of
    # these permissions are needed here.
    source_credentials=source_credentials,
    credential_access_boundary=credential_access_boundary)

# Pass the token directly.
storage_client = storage.Client(
    project='my_project_id', credentials=downscoped_credentials)
# If the source credentials have elevated levels of access, the
# token in flight here will have limited readonly access to objects
# starting with "customer-a" in bucket "bucket-123".
bucket = storage_client.bucket('bucket-123')
blob = bucket.blob('customer-a-data.txt')
print(blob.download_as_string())

Note: Only Cloud Storage supports Credential Access Boundaries. Other Google Cloud services do not support this feature.

Identity Tokens

Google OpenID Connect tokens are available through Service Account, Impersonated, and Compute Engine. These tokens can be used to authenticate against Cloud Functions, Cloud Run, a user service behind Identity Aware Proxy or any other service capable of verifying a Google ID Token.

ServiceAccount

from google.oauth2 import service_account

target_audience = 'https://example.com'

creds = service_account.IDTokenCredentials.from_service_account_file(
        '/path/to/svc.json',
        target_audience=target_audience)

Compute

from google.auth import compute_engine
import google.auth.transport.requests

target_audience = 'https://example.com'

request = google.auth.transport.requests.Request()
creds = compute_engine.IDTokenCredentials(request,
                        target_audience=target_audience)

Impersonated

from google.auth import impersonated_credentials

# get target_credentials from a source_credential

target_audience = 'https://example.com'

creds = impersonated_credentials.IDTokenCredentials(
                                  target_credentials,
                                  target_audience=target_audience)

If your application runs on App Engine, Cloud Run, Compute Engine, or has application default credentials set via GOOGLE_APPLICATION_CREDENTIALS environment variable, you can also use google.oauth2.id_token.fetch_id_token to obtain an ID token from your current running environment. The following is an example

import google.oauth2.id_token
import google.auth.transport.requests

request = google.auth.transport.requests.Request()
target_audience = "https://pubsub.googleapis.com"

id_token = google.oauth2.id_token.fetch_id_token(request, target_audience)

IDToken verification can be done for various type of IDTokens using the google.oauth2.id_token module. It supports ID token signed with RS256 and ES256 algorithms. However, ES256 algorithm won’t be available unless cryptography dependency of version at least 1.4.0 is installed. You can check the dependency with pip freeze or try from google.auth.crypt import es256. The following is an example of verifying ID tokens

from google.auth2 import id_token

request = google.auth.transport.requests.Request()

try:
    decoded_token = id_token.verify_token(token_to_verify,request)
except ValueError:
    # Verification failed.

A sample end-to-end flow using an ID Token against a Cloud Run endpoint maybe

from google.oauth2 import id_token
from google.oauth2 import service_account
import google.auth
import google.auth.transport.requests
from google.auth.transport.requests import AuthorizedSession

target_audience = 'https://your-cloud-run-app.a.run.app'
url = 'https://your-cloud-run-app.a.run.app'

creds = service_account.IDTokenCredentials.from_service_account_file(
        '/path/to/svc.json', target_audience=target_audience)

authed_session = AuthorizedSession(creds)

# make authenticated request and print the response, status_code
resp = authed_session.get(url)
print(resp.status_code)
print(resp.text)

# to verify an ID Token
request = google.auth.transport.requests.Request()
token = creds.token
print(token)
print(id_token.verify_token(token,request))

Making authenticated requests

Once you have credentials you can attach them to a transport. You can then use this transport to make authenticated requests to APIs. google-auth supports several different transports. Typically, it’s up to your application or an opinionated client library to decide which transport to use.

Requests

The recommended HTTP transport is google.auth.transport.requests which uses the Requests library. To make authenticated requests using Requests you use a custom Session object:

from google.auth.transport.requests import AuthorizedSession

authed_session = AuthorizedSession(credentials)

response = authed_session.get(
    'https://www.googleapis.com/storage/v1/b')

urllib3

urllib3 is the underlying HTTP library used by Requests and can also be used with google-auth. urllib3’s interface isn’t as high-level as Requests but it can be useful in situations where you need more control over how HTTP requests are made. To make authenticated requests using urllib3 create an instance of google.auth.transport.urllib3.AuthorizedHttp:

from google.auth.transport.urllib3 import AuthorizedHttp

authed_http = AuthorizedHttp(credentials)

response = authed_http.request(
    'GET', 'https://www.googleapis.com/storage/v1/b')

You can also construct your own urllib3.PoolManager instance and pass it to AuthorizedHttp:

import urllib3

http = urllib3.PoolManager()
authed_http = AuthorizedHttp(credentials, http)

gRPC

gRPC is an RPC framework that uses Protocol Buffers over HTTP 2.0. google-auth can provide Call Credentials for gRPC. The easiest way to do this is to use google-auth to create the gRPC channel:

import google.auth.transport.grpc
import google.auth.transport.requests

http_request = google.auth.transport.requests.Request()

channel = google.auth.transport.grpc.secure_authorized_channel(
    credentials, http_request, 'pubsub.googleapis.com:443')

Note

Even though gRPC is its own transport, you still need to use one of the other HTTP transports with gRPC. The reason is that most credential types need to make HTTP requests in order to refresh their access token. The sample above uses the Requests transport, but any HTTP transport can be used. Additionally, if you know that your credentials do not need to make HTTP requests in order to refresh (as is the case with jwt.Credentials) then you can specify None.

Alternatively, you can create the channel yourself and use google.auth.transport.grpc.AuthMetadataPlugin:

import grpc

metadata_plugin = AuthMetadataPlugin(credentials, http_request)

# Create a set of grpc.CallCredentials using the metadata plugin.
google_auth_credentials = grpc.metadata_call_credentials(
    metadata_plugin)

# Create SSL channel credentials.
ssl_credentials = grpc.ssl_channel_credentials()

# Combine the ssl credentials and the authorization credentials.
composite_credentials = grpc.composite_channel_credentials(
    ssl_credentials, google_auth_credentials)

channel = grpc.secure_channel(
    'pubsub.googleapis.com:443', composite_credentials)

You can use this channel to make a gRPC stub that makes authenticated requests to a gRPC service:

from google.pubsub.v1 import pubsub_pb2

pubsub = pubsub_pb2.PublisherStub(channel)

response = pubsub.ListTopics(
    pubsub_pb2.ListTopicsRequest(project='your-project'))