Scalesec GCP Workload Identity Federation

This package provides a python module to federate access from AWS to GCP using Workload Identity. View our blog for additional details.

Prerequisites

• A GCP service account (environment variable "GCP_SERVICE_ACCOUNT_EMAIL")
• An AWS IAM role (environment variable "AWS_ROLE_NAME")
• AWS credentials (environment variable "AWS_PROFILE")
• python3.x

Quick start

# Create venv and install package
make setup
source .venv/bin/activate
pip install scalesec-gcp-workload-identity

# Rename example .env
mv .env.example .env

# Enter your own environment variables
cat <<EOF >.env
# GCP
export GCP_PROJECT_NUMBER=
export GCP_PROJECT_ID=

# gcp workload identity pool id
export GCP_WORKLOAD_ID=
export GCP_WORKLOAD_PROVIDER=
export GCP_SERVICE_ACCOUNT_EMAIL=

# aws
export AWS_REGION=
export AWS_ACCOUNT_ID=
export AWS_ROLE_NAME=

# Non-required vars
export TOKEN_LIFETIME=
export TOKEN_SCOPES=
EOF

# Source the environment variables so they are exposed
source .env

# set up GCP credentials
gcloud auth login

# Configure the default project
gcloud config set project $GCP_PROJECT_ID

# Enable the STS service in the project
gcloud services enable sts.googleapis.com

# Enable the IAM credentials service
gcloud services enable iamcredentials.googleapis.com

# The following commands use the .env values

# Create the GCP Workload Identity Pool
gcloud beta iam workload-identity-pools create "$GCP_WORKLOAD_ID" \
  --location="global" \
  --description="$GCP_WORKLOAD_ID" \
  --display-name="$GCP_WORKLOAD_ID"

# Create the GCP Workload Identity AWS Provider
gcloud beta iam workload-identity-pools providers create-aws "$GCP_WORKLOAD_PROVIDER" \
  --location="global" \
  --workload-identity-pool="$GCP_WORKLOAD_ID" \
  --account-id="$AWS_ACCOUNT_ID"

# Add the appropriate IAM binding to a pre-existing service account
gcloud iam service-accounts add-iam-policy-binding $GCP_SERVICE_ACCOUNT_EMAIL \
  --role roles/iam.workloadIdentityUser \
  --member "principalSet://iam.googleapis.com/projects/$GCP_PROJECT_NUMBER/locations/global/workloadIdentityPools/$GCP_WORKLOAD_ID/attribute.aws_role/arn:aws:sts::${AWS_ACCOUNT_ID}:assumed-role/$AWS_ROLE_NAME"

Using the module

Set your AWS credentials

export AWS_PROFILE=xyz

Getting a Service Account token is now simple:

from scalesec_gcp_workload_identity.main import TokenService
from os import getenv

# The arguments to TokenService can be ingested
# from the environment if they were exported above.
# Otherwise, pass in your own arguments

token_service = TokenService(
  gcp_project_number=getenv('GCP_PROJECT_NUMBER'),
  gcp_workload_id=getenv('GCP_WORKLOAD_ID'),
  gcp_workload_provider=getenv('GCP_WORKLOAD_PROVIDER'),
  gcp_service_account_email=getenv('GCP_SERVICE_ACCOUNT_EMAIL'),
  aws_account_id=getenv('AWS_ACCOUNT_ID'),
  aws_role_name=getenv('AWS_ROLE_NAME'),
  aws_region=getenv('AWS_REGION'),
  gcp_token_lifetime=getenv('TOKEN_LIFETIME'), # Not required
  gcp_token_scopes=getenv('TOKEN_SCOPES') # Not required
)

sa_token, expiry_date = token_service.get_token()

Token expiration

The default expiration for a service account token is 1h in GCP. This behavior can be changed by overriding the environment variable TOKEN_LIFETIME in the .env file. By default, GCP does not allow tokens to have an expiry over 1 hour and an organization policy must be updated for this change to take affect. The organization policy is called iam.allowServiceAccountCredentialLifetimeExtension and it accepts a list of service accounts that are allowed to have an > 1 hr token.

# To configure the organization policy
gcloud org-policies set-policy policy.yaml

# An example policy.json:
name: projects/1234567890/policies/iam.allowServiceAccountCredentialLifetimeExtension
spec:
  etag: BwXBMNmIrQg=
  rules:
  - values:
      allowedValues:
      - your-sa@yourproject.iam.gserviceaccount.com

Token scopes

The default scope for the service account token is https://www.googleapis.com/auth/cloud-platform. This behaviour can be overridden to enable a different set of scopes by using the environment variable TOKEN_SCOPES in the .env file with a comma-separated list of GCP scopes.

Testing

# make a venv
make setup

Edit .env with your values

# install deps
make dev

# run pytest
make test

Local Linting

To test that your code will pass the lint and code quality GitHub action:

• Clone the repository locally
• Make your updates
• From the root of the repository, execute:

pylint --rcfile .github/workflows/configs/.pylintrc scalesec_gcp_workload_identity tests examples

Examples

We have provided examples on how to use the service account access token generated by this module. Access tokens are mainly used via an API call or using curl on the CLI.

Restricting Identity Pool Providers

By default, any GCP user with the roles/iam.workloadIdentityPoolAdmin or roles/owner role is able to create a workload identity pool in your GCP organization. There are two organization policies available to help you lockdown which outside providers can have pools in your organization.

• constraints/iam.workloadIdentityPoolProviders - Accepts a list of URIs such as https://sts.amazonaws.com or https://sts.windows.net/$AZURE_TENANT_ID. For example:

# Allows all AWS accounts but no Azure or OIDC
gcloud beta resource-manager org-policies allow constraints/iam.workloadIdentityPoolProviders \
     https://sts.amazonaws.com --organization=$ORG_ID

# Allows only a specific Azure tenant but no AWS or OIDC
gcloud beta resource-manager org-policies allow constraints/iam.workloadIdentityPoolProviders \
     https://sts.windows.net/$AZURE_TENANT_ID --organization=$ORG_ID

• constraints/iam.workloadIdentityPoolAwsAccounts - Specifically focused on AWS, this constraint accepts a list of AWS account IDs. If this orgnanization policy is used, constraints/iam.workloadIdentityPoolProviders must either allow https://sts.amazonaws.com or be set to default (allow all). For example:

# Only allows a specific AWS account
gcloud beta resource-manager org-policies allow constraints/iam.workloadIdentityPoolAwsAccounts \
    $AWS_ACCOUNT_ID --organization=$ORG_ID

Upload to PyPi

Set your token/credentials in ~/.pypirc

make dist VERSION=1.x.x

Feedback

Feedback is welcome and encouraged via a GitHub issue. Please open an issue for any bugs, feature requests, or general improvements you would like to see. Thank you in advance!