Flask Azure AD OAuth Provider

Python Flask extension for securing apps with Azure Active Directory OAuth

Purpose

Provide an AuthLib Resource Protector/Server to authenticate and authorise users and applications using a Flask application with OAuth functionality offered by Azure Active Directory, as part of the Microsoft identity platform.

Azure Active Directory, acting as an identity provider, issues OAuth access tokens, the claims of which are validated by this provider. These claims include the identity of the user and client application (used for authentication), and any permissions/scopes assigned or delegated to the user or application (used for authorisation).

This provider supports these scenarios:

1. application to application
   • supports authentication and authorisation
   • used to allow a client application access to some functionality or resources provided by another application
   • can be used for non-interactive, machine-to-machine, processes (using the OAuth Client Credentials Grant)
   • optionally, uses the identity of the client application for authentication
   • optionally, uses permissions assigned directly to the client application for authorisation
2. user to application
   • supports authentication and authorisation
   • used to allow users access to some functionality or resources provided by another application
   • can be used for interactive console (using the Device Authorization Grant) or web application (using the OAuth Authorization Code Grant) processes
   • uses the identity of the user, and optionally, the client application they are using, for authentication
   • optionally, uses permissions assigned to the user, permissions delegated by the user to the client application, and/or permissions assigned directly to the client application for authorisation

Other scenarios may work but are not officially supported, this may change in the future.

Note: This provider does not support client applications requesting tokens from Azure. See the Microsoft Authentication Library (MSAL) for Python package if you need to do this.

Note: This provider has been written to solve an internal need within applications used by the British Antarctic Survey. It is offered to others in the hope that's useful for your needs as well, however it does not (and cannot) cover every option available.

Installation

This package can be installed using Pip from PyPi:

$ pip install flask-azure-oauth

Note: Since version 0.6.0, this package requires Flask 2.0 or greater.

Usage

This provider provides an AuthLib Resource Protector which can be used as a decorator on Flask routes.

A minimal application would look like this:

from flask import Flask

from flask_azure_oauth_theblog_cs2022 import FlaskAzureOauth

app = Flask(__name__)

app.config['AZURE_OAUTH_TENANCY'] = 'xxx'
app.config['AZURE_OAUTH_APPLICATION_ID'] = 'xxx'

auth = FlaskAzureOauth()
auth.init_app(app)


@app.route('/unprotected')
def unprotected():
    return 'hello world'


@app.route('/protected')
@auth()
def protected():
    return 'hello authenticated entity'


@app.route('/protected-with-single-scope')
@auth('required-scope')
def protected_with_scope():
    return 'hello authenticated and authorised entity'


@app.route('/protected-with-multiple-scopes')
@auth('required-scope1 required-scope2')
def protected_with_multiple_scopes():
    return 'hello authenticated and authorised entity'

To restrict a route to any valid user or client application (authentication):

• add the resource protector as a decorator (auth in this example) - for example the /protected route

To restrict a route to specific users (authorisation):

• add any required Scopes to the decorator - for example the /projected-with-* routes

Independently of these options, it's possible to require specific, trusted, client applications, regardless of the user using them. This is useful in circumstances where a user may be authorised but the client can't be trusted:

• set the AZURE_OAUTH_CLIENT_APPLICATION_IDS config option to a list of Azure application identifiers

For example:

app.config['AZURE_OAUTH_CLIENT_APPLICATION_IDS'] = ['xxx']`

Configuration options

The resource protector requires two configuration options to validate tokens correctly. These are read from the Flask config object through the init_app() method.

Configuration Option	Data Type	Required	Description
AZURE_OAUTH_TENANCY	Str	Yes	ID of the Azure AD tenancy all applications and users are registered within
AZURE_OAUTH_APPLICATION_ID	Str	Yes	ID of the Azure AD application registration for the application being protected
AZURE_OAUTH_CLIENT_APPLICATION_IDS	List[Str]	No	ID(s) of the Azure AD application registration(s) for the application(s) granted access to the application being protected
AZURE_B2C_TENANT_MODE	Bool	No	Enables support for Azure AD B2C tenancy, enabling dynamic JWKS source acquisition from userflows (Default: False)
AZURE_TENANT_NAME	Str	No	The name of your Azure AD B2C tenancy, only required when AZURE_B2C_TENANT_MODE is enabled
AZURE_B2C_REGISTERLOGIN_USERFLOW_NAME	Str	No	The name which represents the registration and login userflow in your B2C tenancy. Only required when AZURE_B2C_TENANT_MODE is enabled

Note: If the AZURE_OAUTH_CLIENT_APPLICATION_IDS option is not set, all client applications will be trusted and the azp claim, if present, is ignored.

Before these options can be set you will need to:

1. register the application to be protected
2. define the permissions and roles this application supports
3. register the application(s) that will use the protected application
4. assign permissions to users and/or client application(s)

Flask session support

This provider extends the AuthLib ResourceProtector to support detecting access tokens stored in the Flask session.

This is intended for browser based applications where the Authorization header cannot be easily set to include the access token. This support will be enabled automatically if an access_token session key is set.

Access token versions

Since version 0.5.0, this provider is compatible with Azure access token versions 1.0 and 2.0. Prior to version 0.5.0 only version 2.0 tokens could be used. See Microsoft's documentation for the differences between token versions.

Note: If you use version 1.0 tokens, this provider expects at least one of the identifierUris property values to be api://{protected_application_id}, where {protected_application_id} is the application ID of the app registration representing the application being protected by this provider. Without this, you will receive errors for an invalid audience.

Applications, users, groups and tenancies

Azure Active Directory has a number of different concepts for agents that represent things being protected and things that want to interact with protected things:

• applications - represent services that offer, or wish to use, functionality that should be restricted:
  • services offering functionality are protected applications, e.g. an API
  • services wishing to use functionality interactively or non-interactively, are client applications:
    • interactive client applications include self-service portals for example
    • non-interactive client applications include nightly synchronisation tasks for example
• users - represent individuals that wish to use functionality offered by protected applications, through one or more client applications (e.g. a user may use a self-service portal to access information)
• groups - represent multiple users, for ease of managing permissions to similar users (e.g. administrative users)

For management purposes, all agents are scoped to an Azure tenancy (with the exception of users that can be used across tenancies).

In the Azure management portal:

• applications are represented by Application registrations
• users are represented by users, or optionally groups of users

Permissions, roles and scopes

Azure Active Directory has a number of mechanisms for controlling how agents can interact with each other:

• roles - functions, designations or labels conferred on users and/or groups (e.g. admins, staff)
• direct permissions - capabilities of a protected application client applications can use themselves or without the consent of the current user (e.g. machine-to-machine access to, or modification of, data from all users)
• delegated permissions - capabilities of a protected application the current user allows a client application to use (e.g. interactive access to, or modification of, their data)

Generally, and in terms of the OAuth ecosystem, all of these can be considered as scopes. As discussed in the Usage section, scopes can be used to control who and/or what can use features within protected applications.

Scopes are included the access token generated by a client application (possibly interactively by a user) and presented to the projected application as a bearer token. Azure encodes different mechanisms in different claims:

• roles - for roles assigned to users and permissions directly assigned to client applications
• scp - for permissions delegated by the user to a client application

For ease of use, this extension abstracts these two claims into a single set of scopes that can be required for a given route. Multiple scopes can be required (as a logical AND) to allow scopes to be used more flexibly.

Defining permissions and roles within an application

Permissions and roles are defined in the application manifest of each application being protected. They can then be assigned to users, groups and client applications.

1. register the application to be protected
2. add permissions to application manifest

For example:

"appRoles": [
  {
    "allowedMemberTypes": [
      "Application"
    ],
    "displayName": "List all Foo resources",
    "id": "112b3a76-2dd0-4d09-9976-9f94b2ed965d",
    "isEnabled": true,
    "description": "Allows access to basic information for all Foo resources",
    "value": "Foo.List.All"
  }
],

Assigning permissions and roles to users and applications

Permissions and roles (collectively, application roles) are assigned through the Azure portal:

1. define roles and permissions in the protected application
2. register the client application(s)
3. assign:
   • roles to users/groups
   • permissions to client applications

For assigning permissions:

• permissions can be delegated to client applications, with the agreement of the current user
• permissions can be directly assigned to client applications, with the agreement of a tenancy administrator

Note: Direct assignment is needed for non-interactive applications, such as daemons.

Registering an application in Azure

Follow these instructions.

Note: These instructions apply both to applications that protected by this provider (protected applications), and those that will be granted access to use such applications, possibly by a user (client applications).

Testing support

For testing applications, a local/test JSON Web Key Set (JWKS) can be used to sign local/test JSON Web Tokens (JWTs) without relying on Azure. Local tokens can include, or not include, arbitrary scopes/roles, which can ensure requirements for specific scopes are properly enforced by this provider.

This requires using local tokens signed by the test keys, and patching the FlaskAzureOauth._get_jwks method to validate tokens using the same test keys.

For example:

import unittest

from http import HTTPStatus
from unittest.mock import patch

from flask_azure_oauth_theblog_cs2022 import FlaskAzureOauth
from flask_azure_oauth_theblog_cs2022.mocks.keys import TestJwk
from flask_azure_oauth_theblog_cs2022.mocks.tokens import TestJwt

from examples import create_app


class AppTestCase(unittest.TestCase):
    def setUp(self):
        self.test_jwks = TestJwk()

        with patch.object(FlaskAzureOauth, "_get_jwks") as mocked_get_jwks:
            mocked_get_jwks.return_value = self.test_jwks.jwks()

            # `self.app` should be set to a Flask application, either by direct import, or by calling an app factory
            self.app = create_app()

            self.app.config["TEST_JWKS"] = self.test_jwks
            self.app_context = self.app.app_context()
            self.app_context.push()
            self.client = self.app.test_client()

    def test_protected_route_with_multiple_scopes_authorised(self):
        # Generate token with required roles
        token = TestJwt(
            app=self.app, roles=["BAS.MAGIC.ADD.Records.Publish.All", "BAS.MAGIC.ADD.Records.ReadWrite.All"]
        )

        # Make request to protected route with token
        response = self.client.get(
            "/protected-with-multiple-scopes", headers={"authorization": f"bearer {token.dumps()}"}
        )
        self.assertEqual(HTTPStatus.OK, response.status_code)
        self.app_context.pop()

    def test_protected_route_with_multiple_scopes_unauthorised(self):
        # Generate token with no scopes
        token = TestJwt(app=self.app)

        # Make request to protected route with token
        response = self.client.get(
            "/protected-with-multiple-scopes", headers={"authorization": f"bearer {token.dumps()}"}
        )
        self.assertEqual(HTTPStatus.FORBIDDEN, response.status_code)
        self.app_context.pop()

Developing

This provider is developed as a Python library. A bundled Flask application is used to simulate its usage and act as framework for running tests etc.

Development environment

Git and Poetry are required to set up a local development environment of this project.

Note: If you use Pyenv, this project sets a local Python version for consistency.

# clone from the BAS GitLab instance if possible
$ git clone https://gitlab.data.bas.ac.uk/web-apps/flask-extensions/flask-azure-oauth.git

# alternatively, clone from the GitHub mirror
$ git clone https://github.com/antarctica/flask-azure-oauth.git

# setup virtual environment
$ cd flask-azure-oauth
$ poetry install

Code Style

PEP-8 style and formatting guidelines must be used for this project, except the 80 character line limit. Black is used for formatting, configured in pyproject.toml and enforced as part of Python code linting.

Black can be integrated with a range of editors, such as PyCharm, to apply formatting automatically when saving files.

To apply formatting manually:

$ poetry run black src/ tests/

Code Linting (Python)

Flake8 and various extensions are used to lint Python files in the bas_metadata_library module. Specific checks, and any configuration options, are documented in the ./.flake8 config file.

To check files manually:

$ poetry run flake8 src/ examples/

Checks are run automatically in Continuous Integration.

Dependencies

Python dependencies for this project are managed with Poetry in pyproject.toml.

Non-code files, such as static files, can also be included in the Python package using the include key in pyproject.toml.

Adding new dependencies

To add a new (development) dependency:

$ poetry add [dependency] (--dev)

Then update the Docker image used for CI/CD builds and push to the BAS Docker Registry (which is provided by GitLab):

$ docker build -f gitlab-ci.Dockerfile -t docker-registry.data.bas.ac.uk/web-apps/flask-extensions/flask-azure-oauth:latest .
$ docker push docker-registry.data.bas.ac.uk/web-apps/flask-extensions/flask-azure-oauth:latest

Updating dependencies

$ poetry update

See the instructions above to update the Docker image used in CI/CD.

Dependency vulnerability checks

The Safety package is used to check dependencies against known vulnerabilities.

IMPORTANT! As with all security tools, Safety is an aid for spotting common mistakes, not a guarantee of secure code. In particular this is using the free vulnerability database, which is updated less frequently than paid options.

This is a good tool for spotting low-hanging fruit in terms of vulnerabilities. It isn't a substitute for proper vetting of dependencies, or a proper audit of potential issues by security professionals. If in any doubt you MUST seek proper advice.

Checks are run automatically in Continuous Integration.

To check locally:

$ poetry export --without-hashes -f requirements.txt | poetry run safety check --full-report --stdin

authlib package

The authlib dependency is locked to version 0.14.3 as the 0.15.x release series contains a bug that prevents the kid claim from being accessed from Jason Web Key (JWK) instances. This is a known issue and will be resolved in the 1.x release. See https://github.com/lepture/authlib/issues/314 for more information.

Static security scanning

To ensure the security of this API, source code is checked against Bandit and enforced as part of Python code linting.

Warning: Bandit is a static analysis tool and can't check for issues that are only be detectable when running the application. As with all security tools, Bandit is an aid for spotting common mistakes, not a guarantee of secure code.

To check manually:

$ poetry run bandit -r src/ examples/

Note: This package contains a number of testing methods that deliberately do insecure or nonsensical things. These are necessary to test failure modes and error handling, they are not a risk when using this package as intended. These workarounds have been exempted from these security checks where they apply.

Checks are run automatically in Continuous Integration.

Testing

Integration tests

This project uses integration tests to ensure features work as expected and to guard against regressions and vulnerabilities.

The Python UnitTest library is used for running tests using Flask's test framework. Test cases are defined in files within tests/ and are automatically loaded when using the test Flask CLI command included in the local Flask application in the development environment.

To run tests manually using PyCharm, use the included App (tests) run/debug configuration.

To run tests manually:

$ FLASK_APP=examples FLASK_ENV=testing poetry run python -m unittest discover

Tests are ran automatically in Continuous Integration.

Continuous Integration

All commits will trigger a Continuous Integration process using GitLab's CI/CD platform, configured in .gitlab-ci.yml.

Test/Example applications

For verifying this provider works for real-world use-cases, a test Flask application is included in examples/__init__.py. This test application acts as both an application providing access to, and accessing, protected resources. It can use a number of application registrations registered in the BAS Web & Applications Test Azure AD.

These applications allow testing different versions of access tokens for example. These applications are intended for testing only. They do not represent real applications, or contain any sensitive or protected information.

To test requesting resources from protected resources as an API, set the appropriate config options and run the application container:

$ FLASK_APP=examples poetry run flask

To test requesting resources from protected resources as a browser application, set the appropriate config options and start the application container:

$ FLASK_APP=examples poetry run flask run

Terraform is used to provision the application registrations used:

$ cd provisioning/terraform
$ docker-compose run terraform
$ az login --allow-no-subscriptions
$ terraform init
$ terraform validate
$ terraform apply

Note: Several properties in the application registration resources require setting once the registration has been initially made (identifiers for example). These will need commenting out before use.

Some properties, such as client secrets, can only be set once applications have been registered in the Azure Portal.

Terraform state information is held in the BAS Terraform Remote State project (internal).

Deployment

Python package

This project is distributed as a Python package, hosted in PyPi.

Source and binary packages are built and published automatically using Poetry in Continuous Deployment.

Note: Except for tagged releases, Python packages built in CD will use 0.0.0 as a version to indicate they are not formal releases.

Continuous Deployment

A Continuous Deployment process using GitLab's CI/CD platform is configured in .gitlab-ci.yml.

Release procedure

For all releases:

1. create a release branch
2. bump the version as appropriate in pyproject.toml
3. close release in CHANGELOG.md
4. push changes, merge the release branch into main, and tag with version

The project will be built and published to PyPi automatically through Continuous Deployment.

Feedback

The maintainer of this project is the BAS Web & Applications Team, they can be contacted at: servicedesk@bas.ac.uk.

Issue tracking

This project uses issue tracking, see the Issue tracker for more information.

Note: Read & write access to this issue tracker is restricted. Contact the project maintainer to request access.

License

Copyright (c) 2019-2022 UK Research and Innovation (UKRI), British Antarctic Survey.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.