A production ready Rails Engine that turns your app into an OAuth2 Provider.

• Demo OAuth Provider app with oPRO on Heroku
• Built in oPRO docs
• Example iOS App
• Would you like a Mobile app with that (oPRO tutorial)

oPRO is short for (O)Auth (Pro)vider and is pronounced "oh proh". Not to be confused with Oprah, who does not support or endorse this ruby gem in any way (yet).

Lets say you've built a Rails app, awesome. Now you want to build a mobile app on say, the iPhone... cool. You start throwing around #to_json like nobody's business, but then you realize you need to authenticate users somehow. "Basic Auth!!", you exclaim, but then you realize that's not the most secure solution. You also realize that some users already signed up with Facebook & Twitter so they don't have a username/password combo. What ever shall you do?

Wouldn't it be great if we could have a token exchange where the user goes to a mobile web view and grants permission, and then we return back an auth token just like the big boys (Facebook, Twitter, cough Foursquare cough). With oPRO, we can add this functionality pretty easily. We'll use your existing authentication strategy and provide some integration end points for your clients to use out of the box.

Gemfile

gem 'opro'

Then run

$ bundle install

and don't forget

$ rails g opro:install

This will put a file in initializers/opro.rb, generate some migrations, and add mount_opro_oauth to your routes.

Now we're ready to migrate the database

$ rake db:migrate

This will add Opro::Oauth::AuthGrant and Opro::Oauth::ClientApp to your database. An iPhone app would need to register for a client_id and client_secret before using OAuth as a ClientApp. Once created they could get authorization from users by going through the OAuth flow, thus creating AuthGrants. In other words, a ClientApp has many users through AuthGrants.

Go to initializers/opro.rb and configure your app for your authentication scheme. If you're not using devise, see "Custom Auth" below.

Opro.setup do |config|
  config.auth_strategy = :devise
end

Now in your controllers you can allow OAuth access using the same syntax of the rails before_filter

class UsersController < ApplicationController
  allow_oauth!  :only => [:show]
end

You can also disallow OAuth on specific actions. Disallowing will always over-ride allowing.

class ProductsController < ApplicationController
  disallow_oauth!   :only => [:create]
end

By default, all OAuth access is blacklisted. To whitelist all access, add allow_oauth! to your ApplicationController (this is not recommended). The best practice is to add allow_oauth! or disallow_oauth to each and every controller.

That should be all you need to do to get set up. Congrats, you're now able to authenticate users using OAuth!!

oPRO comes with built in documentation, so if you start your server you can view the docs at http://localhost:3000/oauth_docs. Or you can view the guide on the example app. This guide will walk you through creating your first OAuth client application, giving access to that app as a logged in user, getting an access token for that user, and using that token to access the server as an authenticated user!

Keep in mind that the OAuth client application is associated to the user creating it. If you have multiple administrators for oPRO, be aware that the web interface is currently only useful to the user that created it.

oPRO is simple by default, but easily configurable for a number of common use cases. Check out the options below.

If you're not using devise, you can manually configure your own auth strategy. In the future I plan on adding more auth strategies; ping me or submit a pull request for your desired authentication scheme.

Opro.setup do |config|
  config.login_method             { |controller, current_user| controller.sign_in(current_user, :bypass => true) }
  config.logout_method            { |controller, current_user| controller.sign_out(current_user) }
  config.authenticate_user_method { |controller| controller.authenticate_user! }
end

When a user authenticates with a client, they are automatically granting read permission to any action that you allow_oauth!. Read-only clients are restricted to using GET requests. By default, oPRO will ask users for write permission on a client by the client application. Client apps with :write permission can use all HTTP verbs including POST, PATCH, PUT, DESTROY on any url you whitelist using allow_oauth!.

To remove write permissions, comment out this line in the oPRO initializer:

config.request_permissions = [:write]

You can add custom permissions by adding to the array:

config.request_permissions = [:write, :email, :picture, :whatever]

You can then restrict access using the custom permissions by calling require_oauth_permissions, which takes the same arguments as before_filter:

require_oauth_permissions :email, :only => :index

You can also skip permissions using skip_oauth_permissions. By default, permissions will just check to see if a client has the permission and will allow the action if it is present. If you want to implement custom permission checks, you can write custom methods using the pattern oauth_client_can_#{permission}?. For example, if you were restricting the :email permission, you would create a method:

def oauth_client_can_email?
  # ...
end

The result is expected to be true or false.

For added security, you can require access_tokens to be refreshed by client applications. This will help to mitigate the risk of a leaked access_token and enable an all around more secure system. This security comes at a price, however, since implementing the refresh_token functionality in a client can be more difficult.

By default, refresh tokens are enabled. You can disable them in your application and set the timeout period of the tokens by adding this line to your configuration:

config.require_refresh_within = false

If you disable refresh tokens and then re-enable it you may have authorization grants that do not have a timeout listed, you can keep it like this or you can fix by iterating through all auth grants and setting their access_token_expires_at like this:

Opro::Oauth::AuthGrant.find_each(:conditions => "access_token_expires_at is null") do |grant|
  grant.access_token_expires_at = Time.now + ::Opro.require_refresh_within
  grant.save
end

You may also need to inform clients that they need to update their credentials and start using refresh tokens.

If a client application has a user's password and username/email, they can exchange these for a token. This is much safer than storing the username and password on a local device, but it does not offer the traditional OAuth "Flow". Because of this, all available permissions will be granted to the client application. If you want to disable this feature you can set the configuration below to false:

config.password_exchange_enabled = true

If you have this feature enabled, you can further control what applications can use the feature. Some providers may wish to have "Blessed" client applications that have this ability while restricting all other clients. To accomplish this, you can create a method in your ApplicationController called oauth_valid_password_auth? that accepts a client_id and client_secret and returns true or false based on whether that application can use password auth:

def oauth_valid_password_auth?(client_id, client_secret)
  BLESSED_APP_IDS.include?(client_id)
end

If you are using this password functionality without a supported authorization engine (like devise), you will need to add an additional method that supports validating whether or not a user's credentials are valid. The method for this is called find_user_for_auth and accepts a controller and the parameters. The output is expected to be a user. Add this to your config like you did to the other required methods in the "Custom Auth" section:

config.find_user_for_auth do |controller, params|
  # user = User.find(params[:something])
  # return user.valid_password?(params[:password]) ? user : false
end

If you're authenticating by exchanging something other than a password (such as a facebook auth token), clients can still enable this functionality by setting params[:grant_type] == 'password' in their initial request. You can then use the find_user_for_auth method from above and implement your custom behavior. You can call find_user_for_auth multiple times and the application will try calling each auth method in order. It is suggested that you return from this block early if the params are missing a vital key like this:

config.find_user_for_auth do |controller, params|
  return false if params[:fb_token].blank?
  User.where(:fb_token => params[:fb_token]).first
end

If your API becomes a runaway success and people start abusing your API, you might choose to limit the rate that client applications can access your API. It is common for popular read-only APIs to have an hourly or daily rate limit to help prevent abuse. If you want this type of functionality, you can use oPRO's built in hooks: one to record the number of times a client application has accessed your API and another to let the application know if the Client app has gone over its allotted rate.

To record the number of times an application has accessed your site add this method to your ApplicationController:

def oauth_client_record_access!(client_id, params)
  # implement your rate counting mechanism here
end

Then to let our server know if a given client has reached its limit, add the method below. The output is expected to be true if the client has gone over their limit and false if they have not:

def oauth_client_rate_limited?(client_id, params)
  # implement your own custom rate limiting logic here
end

Rate limited clients will receive an "unsuccessful" response to any query with a message letting them know they've been rate limited. Using redis with a rotating key generator based on (hour, day, etc.) is one very common way to count accesses and implement the rate limits. Since there are so many different ways to implement this, we decided to give you a blank slate and let you implement it however you would like. The default is that apps are not rate limited, and in general unlimited API access is the way to go, but if you do find abusive behavior you can always easily add in a rate limit.

By default, oPRO allows clients to send their authorization token in a header. For example, someone using an auth token of 9693accessTokena7ca570bbaf could set the Authorization header in a request like this:

    $ curl -H "Authorization: Bearer 9693accessTokena7ca570bbaf" "http://localhost:3000/oauth_test/show_me_the_money"

By default, oPRO will accept Bearer and token in the authorization header, but if your client needs to send a custom auth header, you can add a custom extra regular expression to parse and return the token. For example, if a client was setting the auth header like this:

    $ curl -H "Authorization: cUsTomAuthHeader 9693accessTokena7ca570bbaf" "http://localhost:3000/oauth_test/show_me_the_money"

You could pull out the auth token using the regular expression /cUsTomAuthHeader\s(.*)/. If you're not great with regular expressions, I highly recommend using Rubular to test regex matches. It is very important that we are "capturing" data in between the () characters. The data returned inside of the parens is expected to be the auth token with no spaces or special characters such as new lines or quotes. To parse this auth header in oPRO, you can specify the header_auth_regex in an initializer like this:

Opro.setup do |config|
  config.auth_strategy = :devise

  config.header_auth_regex = /cUsTomAuthHeader\s(.*)/
end

Now when a client sends your custom auth header, it will be parsed correctly. Custom authorization headers should not be used for security through obscurity. They may be exposed in the docs or tests in a later iteration of oPRO. If you have strong feelings against this, then please open a pull request or send me a message stating your case.

By default oPRO ships with views and controllers much like Devise. You can use the built in oPRO views, and change styles: all oPRO views should be wrapped with an opro class you can use for styling. For more control you can override oPRO views with your own views. To do this you will need to provide your own routes/controllers/views. You can override oauth_docs_controller, oauth_docs_controller, and oauth_client_apps_controller.

Currently this is a manual process to give you control and understanding of what is happening behind the scenes, if you want to work on a generator, I will be happy to help you...ping me @schneems. To make things a little more clear we are going to be referencing the oPRO demo rails app.

To start out overriding a controller we need to specify the new controller in your routes inside of mount_opro_oauth, for example if you wanted to over-ride the oauth_client_apps controller with a controller in app/controllers/oauth/client_apps_controller.rb you could specify it like this:

mount_opro_oauth :controllers => {:oauth_client_apps => 'oauth/client_apps'}

You can see an example of setting the routes in the oPRO demo rails app. Of course you then need to create the controller, it needs to inherit from the original controller Opro::Oauth::ClientAppController like this:

class Oauth::ClientAppsController < Opro::Oauth::ClientAppController
end

You can see an example of a: custom oPRO controller. Once you've got your controller finished, you need to specify your custom oPRO views.

It may help to look at the current oPRO controllers and current oPRO views. Again, if you get stuck take a look at the oPRO demo rails app.

If you do not wish for test, docs, or client_apps views & controllers to be available, you can skip them using except in your mount_opro_oauth like this:

mount_opro_oauth :except => [:oauth_client_apps]

We recommend against doing this, but we aren't your mother.

• You have a user model and that is what you're authenticating
• You're using Active Record

If you have a question file an issue or find me on the Twitters @schneems. Another good library for turning your app into an OAuth provider is Doorkeeper, if this project doesn't meet your needs let me know why and use them :)

This project rocks and uses MIT-LICENSE.