What's Hawk and how to use it?

At Mozilla, we recently had to implement the Hawk authentication scheme for a number of projects, and we came up creating two libraries to ease integration into pyramid and node.js apps.

But maybe you don't know Hawk.

Hawk is a relatively new technology, crafted by one of the original OAuth specification authors, that intends to replace the 2-legged OAuth authentication scheme using a simpler approach.

It is an authentication scheme for HTTP, built around HMAC digests of requests and responses.

Every authenticated client request has an Authorization header containing a MAC (Message Authentication Code) and some additional metadata, then each server response to authenticated requests contains a Server-Authorization header that authenticates the response, so the client is sure it comes from the right server.

Exchange of the hawk id and hawk key

To sign the requests, a client needs to retrieve a token id and a token key from the server.

Hawk itself does not define how these credentials should be exchanged between the server and the client. The excellent team behind Firefox Accounts put together a scheme to do that, which acts like the following:

Note

All this derivation crazyness might seem a bit complicated, but don't worry, we put together some libraries that takes care of that for you automatically.

If you are not interested into these details, you can directly jump to the next section to see how to use the libraries.

When your server application needs to send you the credentials, it will return it inside a specific Hawk-Session-Token header. This token can be derived to split this string in two values (hawk id and hawk key) that you will use to sign your next requests.

In order to get the hawk credentials, you'll need to:

First, do an HKDF derivation on the given session token. You'll need to use the following parameters:

key_material = HKDF(hawk_session, "", 'identity.mozilla.com/picl/v1/sessionToken', 32*2)

Note

The identity.mozilla.com/picl/v1/sessionToken is a reference to this way of deriving the credentials, not an actual URL.

Then, the key material you'll get out of the HKDF need to be separated into two parts, the first 32 hex caracters are the hawk id, and the next 32 ones are the hawk key.

Credentials:

credentials = {
    'id': keyMaterial[0:32],
    'key': keyMaterial[32:64],
    'algorithm': 'sha256'
}

Httpie

To showcase APIs in the documentation, I like to use httpie, a curl-replacement with a nicer API, built around the python requests library.

Luckily, HTTPie allows you to plug different authentication schemes for it, so I wrote a wrapper around mohawk to add hawk support to the requests lib.

Doing hawk requests in your terminal is now as simple as:

$ pip install requests-hawk httpie
$ http GET localhost:5000/registration --auth-type=hawk --auth='id:key'

In addition, it will help you to craft requests using the requests library:

import requests
from requests_hawk import HawkAuth

hawk_auth = HawkAuth(
  credentials={'id': id, 'key': key, 'algorithm': 'sha256'})

requests.post("/url", auth=hawk_auth)

Alternatively, if you don't have the token id and key, you can pass the hawk session token I talked about earlier and the lib will take care of the derivation for you:

hawk_auth = HawkAuth(
    hawk_session=resp.headers['hawk-session-token'],
    server_url=self.server_url
)
requests.post("/url", auth=hawk_auth)

Integrate with python pyramid apps

If you're writing pyramid applications, you'll be happy to learn that Ryan Kelly put together a library that makes Hawk work as an Authentication provider for them. I'm chocked how simple it is to use it.

Here is a demo of how we implemented it for Daybed:

from pyramid_hawkauth import HawkAuthenticationPolicy

policy = HawkAuthenticationPolicy(decode_hawk_id=get_hawk_id)
config.set_authentication_policy(authn_policy)

The get_hawk_id function is a function that takes a request and a tokenid and returns a tuple of (token_id, token_key).

How you want to store the tokens and retrieve them is up to you. The default implementation (e.g. if you don't pass a decode_hawk_id function) decodes the key from the token itself, using a master secret on the server (so you don't need to store anything).

Integrate with node.js Express apps

We had to implement Hawk authentication for two node.js projects and finally came up factorizing everything in a library for express, named express-hawkauth.

In order to plug it in your application, you'll need to use it as a middleware:

var express = require("express");
var hawk = require("express-hawkauth");
app = express();

var hawkMiddleware = hawk.getMiddleware({
  hawkOptions: {},
  getSession: function(tokenId, cb) {
    // A function which pass to the cb the key and algorithm for the
    // given token id. First argument of the callback is a potential
    // error.
    cb(null, {key: "key", algorithm: "sha256"});
  },
  createSession: function(id, key, cb) {
    // A function which stores a session for the given id and key.
    // Argument returned is a potential error.
    cb(null);
  },
  setUser: function(req, res, tokenId, cb) {
    // A function that uses req and res, the hawkId when they're known so
    // that it can tweak it. For instance, you can store the tokenId as the
    // user.
    req.user = tokenId;
  }
});

app.get("/hawk-enabled-endpoint", hawkMiddleware);

If you pass the createSession parameter, all non-authenticated requests will create a new hawk session and return it with the response, in the Hawk-Session-Token header.

If you want to only check a valid hawk session exists (without creating a new one), just create a middleware which doesn't have any createSession parameter defined.

Some reference implementations

As a reference, here is how we're using the libraries I'm talking about, in case that helps you to integrate with your projects.