Summary
This article explains how to create an oauth2 custom authorizer for amazon’s AWS API Gateway.
I wanted to use the oauth2 client credentials grant, also known as 2-legged oauth 2 workflow, see: http://oauthbible.com/#oauth-2-two-legged. This kind of workflow is useful for machine to machine communication, where the client machine is also the resource owner.
+---------+ +---------------+
| | | |
| |>--(A)- Client Authentication --->| Authorization |
| Client | | Server |
| |<--(B)---- Access Token ---------<| |
| | | |
+---------+ +---------------+
Figure 6: Client Credentials Flow
Ref: https://tools.ietf.org/html/draft-ietf-oauth-v2-31#section-4.4
If you struggle to workout which grant type use, this diagram can be useful:
Ref: https://oauth2.thephpleague.com/authorization-server/which-grant/
To implement oauth in api gateway we need to carry out the following tasks, which are covered in detail later on:
- Setup an oauth server
- Create a custom authorizer
- Configure the API gateway
Setup and configure an oauth server
The first task was to evaluate what software I could use to act as an authorization and resource server. In order that the custom lambda authorizer could validate a token, I needed an implementation to expose a token validation endpoint as well as the normal token creation endpoint.
Below is a list of candidates I looked at
| Software |
Language |
Description |
| Hydra |
Go |
Opensource. Good documentation. Responsive maintainer(s), PR merged same day. API for token validation |
| PHP OAuth 2.0 Server |
PHP |
Opensource. Good documentation. Unsure if it can validate tokens via an api? |
| Spring Security OAuth |
Java |
Opensource. Good documentation. API for token validation |
I also took a quick look at some other implementations, see: https://oauth.net/code/. In this article I choose to use hydra mainly because i’m familiar with Go, it supported token verification using an api and looked really straight forward to setup and configure
Setting up hydra
I ran hydra using the published docker image: https://hub.docker.com/r/oryam/hydra/
docker run -d --name hydra \
-p 4444:4444 \
-e SYSTEM_SECRET='3bu>TMTNQzMvUtFrtrpJEMsErKo?gVuW' \
-e FORCE_ROOT_CLIENT_CREDENTIALS='8c97eaed-f270-4b2f-9930-03f85160612a:MxGdwYBLZw7qFkUKCFQUeNyvher@jpC]' \
-e HTTPS_TLS_CERT_PATH=/server.crt \
-e HTTPS_TLS_KEY_PATH=/key.pem \
-v $(pwd)/server.crt:/server.crt \
-v $(pwd)/key.pem:/key.pem oryam/hydra
This sets up hydra to use ssl and seeds the root credentials, which is used later on perform administrative tasks with hydra
Self-sign ssl certificate
The api gateway lamda authentication function will need to communicate with the hydra. I choose to secure this communication using SSL/TLS. If you don’t have an SSL certificate for your hydra instance, you could buy one or you can create your own self-signed certificate (for internal usage or test purposes). I choose to go down the self-signed root.
- Create a private key:
openssl genrsa 2048 > key.pem
- Create a signing request:
openssl req -new -key key.pem -out cert.csr
Example answers:
Country Name (2 letter code) [AU]:GB
State or Province Name (full name) [Some-State]:London
Locality Name (eg, city) []:London
Organization Name (eg, company) [Internet Widgits Pty Ltd]:My Company Limited
Organizational Unit Name (eg, section) []:
Common Name (e.g. server FQDN or YOUR name) []:oauth.mycompany.local
Email Address []:
- Sign the request:
openssl x509 -req -days 3650 -in cert.csr -signkey key.pem -out server.crt
You can now use key.pem and server.crt to run the docker container as above.
Configuring hydra
- Create a system token
- Create a client(s)
- Assign policies to a client (optional)
- Test creating a client token
- Test validating a client token
- Test validating a client token against a policy
- Health check endpoint
Create a system token
System token is used to perform administrative interactions with hydra, such as creating clients, validating tokens etc. In the -u (username:password) this is the system client id and secret set in the FORCE_ROOT_CLIENT_CREDENTIALS environment variable
Request
curl -k -X POST \
-d grant_type=client_credentials \
-d scope='hydra hydra.clients' \
https://oauth.mycompany.local/oauth2/token
Result
{
"access_token": "fIyy-W3j2cmNSP40GK9HmQ9wlmhzFpdcxia64JHN3po.ww3Ob46pPaj1tz_XfXG80BAnLy5XbwuLqSjmwnqh6Ks",
"expires_in": 3599,
"scope": "hydra hydra.clients",
"token_type": "bearer"
}
Create a client
Create client, this is a user of your api
Request
curl -k -X POST \
-H 'Authorization: bearer fIyy-W3j2cmNSP40GK9HmQ9wlmhzFpdcxia64JHN3po.ww3Ob46pPaj1tz_XfXG80BAnLy5XbwuLqSjmwnqh6Ks' \
-d '{"id":"3094A219-52B1-4900-91F7-514C4392D8C3","client_name":"Client1","grant_types":["client_credentials"],"response_types":["code"],"public":false}' \
https://oauth.mycompany.local/clients
Note: In the request authorization header we use the access_token we obtained from the previous step. We also specify the client will access the system using the client_credentials grant.
Result
{
"id": "3094A219-52B1-4900-91F7-514C4392D8C3",
"client_name": "Client1",
"client_secret": "(SDk!*ximQS*",
"redirect_uris": null,
"grant_types": [
"client_credentials"
],
"response_types": [
"code"
],
"scope": "",
"owner": "",
"policy_uri": "",
"tos_uri": "",
"client_uri": "",
"logo_uri": "",
"contacts": null,
"public": false
}
Note: The client_secret has been generated. client_id and client_secret are used by the client to (create tokens)[]
Create a policy
In our example we are creating two types of clients, read only and write clients. The curl command below defines the read policy and associates it with subjects, which in the context of hydra, are a comma separated list of client ids.
Request
curl -k -X POST -H \
'Authorization: bearer fIyy-W3j2cmNSP40GK9HmQ9wlmhzFpdcxia64JHN3po.ww3Ob46pPaj1tz_XfXG80BAnLy5XbwuLqSjmwnqh6Ks' \
-d '{"description":"Api readonly policy.","subjects":["3094A219-52B1-4900-91F7-514C4392D8C3"],"actions":["read"],"effect":"allow","resources":["resources:orders:<.*>"]}' \
https://oauth.mycompany.local/policies
Note: In the above request we specify which resources the policy applies to. In our example we are specifying that the orders resource is having a policy applied. The “ is a wild card identifier which applies the policy to orders and any sub resources.
Response
{
"id": "03c59a92-1fa6-4df9-ad1e-e5d551bc2c71",
"description": "Api readonly policy.",
"subjects": [
"3094A219-52B1-4900-91F7-514C4392D8C3"
],
"effect": "allow",
"resources": [
"resources:orders:<.*>"
],
"actions": [
"read"
],
"conditions": {}
}
Create a client token
This call is issued by the client application
Request
curl -k -X POST \
-d grant_type=client_credentials \
-u '3094A219-52B1-4900-91F7-514C4392D8C3:(SDk!*ximQS*' \
https://oauth.mycompany.local/oauth2/token
Result
{
"access_token": "1z4Bb_r8lgmUKaD1FyOgP0tBJ_UIafhX2-QyIvUgLN8.NHdZ3zm4Ly6mepP7flGJQMN6-YfKox3OyPPZiiMg-mk",
"expires_in": 3599,
"scope": "",
"token_type": "bearer"
}
Test validating a client token
This is the call that the lambda function will need to make to validate a client token
Request
curl -k -X POST \
-H 'Authorization: bearer fIyy-W3j2cmNSP40GK9HmQ9wlmhzFpdcxia64JHN3po.ww3Ob46pPaj1tz_XfXG80BAnLy5XbwuLqSjmwnqh6Ks' \
-d 'token=1z4Bb_r8lgmUKaD1FyOgP0tBJ_UIafhX2-QyIvUgLN8.NHdZ3zm4Ly6mepP7flGJQMN6-YfKox3OyPPZiiMg-mk' \
https://oauth.mycompany.local/oauth2/introspect
Response
{"active":true,"client_id":"Client1","sub":"Client1","exp":1481975503,"iat":1481971902,"aud":"Client1"}
Test validating a client token against a policy
This call could be used by the lambda function as an alternative, perhaps mapping the http verb to the policy, i.e GET=read or POST=write
Request
curl -X POST -k \
-H 'Authorization: bearer fIyy-W3j2cmNSP40GK9HmQ9wlmhzFpdcxia64JHN3po.ww3Ob46pPaj1tz_XfXG80BAnLy5XbwuLqSjmwnqh6Ks' \
-d '{"token":"1z4Bb_r8lgmUKaD1FyOgP0tBJ_UIafhX2-QyIvUgLN8.NHdZ3zm4Ly6mepP7flGJQMN6-YfKox3OyPPZiiMg-mk","subject":"Client1","action":"read","resource":"resources:orders:123"}' https://oauth.mycompany.local/warden/token/allowed
Response – Allowed
{"sub":"Client1","scopes":[],"iss":"hydra.localhost","aud":"Client1","iat":"2016-12-17T10:51:42.917937398Z","exp":"2016-12-17T11:51:43.049266177Z","ext":null,"allowed":true}
Health check endpoint
This call is useful for a loadbalancer ALB or ELB to determine if a node is active or not.
Request
curl -k https://oauth.mycompany.local/health -i
Response
Note that there is no content with this response, which is why I included the -i curl parameter to show the response code
HTTP/1.1 204 No Content
Date: Mon, 19 Dec 2016 10:04:59 GMT
Create a custom authorizer
Next up we examine how to create the lambda function to call our hydra server. The code for this authorizer can be found on github: https://github.com/ewilde/oauth2-api-gateway. I used the serverless framework to help me build and deploy the authorizer and test endpoint
Note: This is the first node application i’ve written, so apologies if it’s not very idiomatic.
serverless.yml
functions:
hello:
handler: functions/handler.hello
events:
- http:
path: hello
authorizer: auth
method: get
vpc:
securityGroupIds:
- sg-575c752a
subnetIds:
- subnet-35561a7c
- subnet-4e44d315
- subnet-454bd668
auth.js
I wrote a simple javascript library to interact with hydra which we instantiate here to use later on when validating an incoming token.
var HydraClient = require('./hydra');
var client = new HydraClient();
Below is the function to create the policy document to return to the api gateway when a client presents a valid token
const generatePolicy = (principalId, effect, resource) => {
const authResponse = {};
authResponse.principalId = principalId;
if (effect && resource) {
const policyDocument = {};
policyDocument.Version = '2012-10-17';
policyDocument.Statement = [];
const statementOne = {};
statementOne.Action = 'execute-api:Invoke';
statementOne.Effect = effect;
statementOne.Resource = resource;
policyDocument.Statement[0] = statementOne;
authResponse.policyDocument = policyDocument;
}
return authResponse;
};
Below is the actual authorization method that is called by the api gateway. It validates the incoming request and returns either:
- Error: Invalid token
- Unathorized
- Success: policy document
module.exports.auth = (event, context) => {
client.validateTokenAsync({
'access_token': event.authorizationToken
}, function (result) {
if (result == null) {
console.log(event.authorizationToken + ': did not get a result back from token validation');
context.fail('Error: Invalid token');
} else if (!result.active) {
console.log(event.authorizationToken + ': token no longer active');
context.fail('Unauthorized');
} else {
console.log(event.authorizationToken + ': token is active will allow.');
console.log('principle: ' + result.client_id + ' methodArm: ' + event.methodArn);
var policy = generatePolicy('user|' + result.client_id, 'allow', event.methodArn);
console.log('policy: ' + JSON.stringify(policy));
context.succeed(policy);
}
});
};
hydra client
The function below calls hydra to make sure the token is valid and that the TTL has not expired
function validateToken(systemToken, clientToken, callback) {
var tokenParsed = clientToken.access_token.replace('bearer ', '');
console.log('Validating client token:' + tokenParsed);
var request = require('request');
request.post(
{
url: constants.base_auth_url + '/oauth2/introspect',
agentOptions: {
ca: constants.self_signed_cert
},
headers: {
'Authorization' : 'bearer ' + systemToken.access_token
},
form: {
token: tokenParsed
}
},
function (error, response, body) {
if (!error && response.statusCode >= 200 && response.statusCode < 300) {
var result = JSON.parse(body);
callback(result);
}
else {
console.log(response);
console.log(body);
console.log(error);
callback(null);
}
});
}
##Configuring the API gateway and testing the application
Because we used serverless in this example there is really nothing to be done here. The serverless.yml configures the authorizer:
authorizer: auth for the endpoint /hello
To test the application:
- Deploy the serverless application `serverless deploy’
- Create a token
- In postman make a call to ‘/hello’ passing in your token in the
Authorization header
Useful references
Podcasts
Alex Bilbie – OAuth 2 and API Security
Covers different grant types and what they’re each appropriate for, as well as discuss some potential API security strategies for one of Adam’s personal projects.
http://www.fullstackradio.com/4
Thought machine – API Gateway and lambda
Interesting discussion on lambda architectures
http://martinfowler.com/articles/serverless.html
Interview with Mike Roberts discussing serverless architectures
https://softwareengineeringdaily.com/2016/08/23/serverless-architecture-with-mike-roberts/
Edward Wilde 