Skip to content

Latest commit

 

History

History
193 lines (137 loc) · 6.95 KB

README.md

File metadata and controls

193 lines (137 loc) · 6.95 KB
Raw

Flask-OPA

Build Status codecov PyPI Version

Simple to use Flask extension that lets you secure your projects with Open Policy Agent. It allows

  • HTTP API Authorization
  • Policy Enforcement Point (AOP using decorators on methods)

Quick start

Its recommended for you to try out the app in the package examples. Thanks to the Makefile you can run the demo project with the following command

 make demo

How it works?

For a better understanding of what make demo does and how, you should set up flask_opa in your project. Follow the next steps:

  1. Run OPA in server mode

    • Check the latest OPA release and download it.
    • Put the binary file in the path of your system
    • Allow its execution with something like chmod 755 ./opa
    • Run opa in server mode with the sample policies
    opa run -s -w examples
    • -s is to run it in server mode instead of opening the REPL
    • -w is for watching the changes of the data/policy files

  2. Specify the configuration variables

    • OPA_URL url accessible in your running OPA server, used to evaluate your input. It includes the path of the policy, e.g. http://localhost:8181/v1/data/examples/allow.

    • OPA_SECURED boolean to specify if OPA will be enabled to your application.

    See more at the rest api reference

  3. Bind the OPA class to your Flask application

    It is easy to bind the Flask-OPA library to your application. Just follow the following steps:

    1. Create the OPA instance

      app = Flask(__name__)
app.config.from_pyfile('app.cfg')
opa = OPA(app, parse_input)

      Let's see the parameters that we passed to the OPA class:

      • parse_input (Required) contains a method that returns the input data json to be evaluated by the policy, e.g.:
      {
    "input": {
      "method": "GET",
      "path": ["data", "jon"],
      "user": "paul"
    }
}
      • url (Optional) to use an specific url instead of the OPA_URL optionally specified in the app configuration.
      • allow_function (Optional) predicate that determinate if the response from OPA allows (True) or denies (False) the request

      If you want enforce the OPA security in your application you can create the OPA instance like this:

      opa = OPA.secure(app, parse_input, url="http://localhost:8181/v1/data/package_name/allow")

      or

      opa = OPA(app, parse_input, url="http://localhost:8181/v1/data/package_name/allow").secured()

      otherwise, OPA will enforce your security only if OPA_SECURED is True.

      Specify the logging level to DEBUG if you want to get access to Flask-OPA logs of its operations using

      app.logger.setLevel(logging.DEBUG)

    2. Run your Flask application.

Policy Enforcement point

One of the features this module provides is Policy Enforcement Point, which basically allows you to ensure policies at any method of your application. For practical purposes, lets imagine a sample method that is in charge of logging content related to some actions done by users. In this case we must create a different input functions that provide useful information for certain policies that will decide if a log should be sent or not to a remote server. Let's suppose that such logging method is something like:

def log_remotely(content):
    # Imagine a code to log this remotely
    app.logger.info("Logged remotely: %s", content)

Let's create a PEP decorator using our OPA instance as a function (callable mode) that will intercept every call to log_remotely. The parameters are pretty much the same as those used to secure the application. The resulting instance will decorate our function of interest:

def validate_logging_input_function(*arg, **kwargs):
   return {
        "input": {
            "user": request.headers.get("Authorization", ""),
            "content": arg[0]
        }
    }

secure_logging = app.opa("Logging PEP", app.config["OPA_URL_LOGGING"], validate_logging_input_function)

@secure_logging
def log_remotely(content):
    # Imagine a code to log content remotely
    app.logger.info("Logged remotely: %s", content)

As you might have noticed, the only new thing we truly require for adding the PEP is a new input function. This function can provide a more versatile input than the one used by the OPA instance created for the whole app: in our example it provides data related to the user request and data provided by the parameters of the decorated function as well.

Read the examples README for more detailed information about how to run a demo.

Error handling

All errors related to OPA extend from OPAException. They will always be thrown unless the app variable OPA_DENY_ON_FAIL or app.opa.deny_on_opa_fail is set to False.

Types of OPAException errors

  • AccessDeniedException: When the allow_function returns False, indicating that a policy denies the access.
  • OPAServerUnavailableException: When it cannot connect to the OPA Server.
  • OPAUnexpectedException: When the response of the OPA server is not OK, i.e. the status code is not 200.

Handling OPA Exceptions

With the errorhandler decorator of the Flask app, you can easily catch any of these errors, e.g.:

@app.errorhandler(OPAException)
def handle_opa_exception(e):
    return json.dumps({"message": str(e)}), 403

or particular ones:

@app.errorhandler(OPAServerUnavailableException)
def handle_opa_exception_conn(e):
    app.logger.debug("Issue connecting to the OPA server: %s", e)
    return "Authorization cannot be enforced", 403

Makefile

The Makefile contains multiple useful actions you might need. Check them with

 make help

Author

Eliecer Hernandez Garbey

Links

License

This project is licensed under the MIT License - see the LICENSE.md file for details.