OAuth 2.0 and OpenID Connect
OAuth 2.0 is for authorization while OpenID Connect is for authentication.
OAuth 2.0 is a delegated authorization services. At a high level, it works as follows:
- An application will integrate a “Connect with website_name” button
- On click, the user browser will redirect over to that website and they will be prompted to login (email and password). With the redirection, the callback URL, response type and scope(s) are specified
- After successful login, the website will ask something along the line of “Allow application to access whatever_info?” with a yes/no call to action. This is important: explicit consent is key
- If the user click “yes”, the browser redirects back to the first application callback or redirect URI with an authorization code.
- The application use the authorization code to ask the website for an access token.
- The website check the authenticity and validity of the authorization token and exchange it with an access token.
- Now the application contacts the website’s API and ask for whatever_info by presenting the authorization token thing that let the website knows the application is allowed to access this information.
OAuth is a authorization service. However, starting with Google or Facebook login button, it also became an authentication service, which it wasn’t designed for at first. This implementation introduces problem because there is no standard way to get user’s information and every implementation and scopes are different.
OAuth 2.0 uses confusing terminologies that just renames things that are already named.
- Resource owner: you and me (the user who clicks on “yes” or “no” and who owns the data)
- Client: the application that want to access the data
- Authorization server: the website the user is redirected to allow the sharing of the data
- Resource server: the server with the API or system that actually holds the data (not always the same as the authorization server)
- Authorization grant: the “special magic thing” to prove the user consented to share the data
- Redirect URI: where the browser should go back at the end of the consent dance executed beforehand
- Access token: the actual token or key needed to access the data for which permission was granted, on the resource server
- Scopes/Consents: a list of scopes that the authorization server understands. A list of permissions that make sense for this system
The reason there is two tokens (authorization grant and access token) is because of network security concepts called back channel (highly secure channel) and front channel (less secure channel).
Most of the OAuth ceremony occurs on the back channel, i.e., between the browser and a website. In the case where there is only an authorization grant, and if the network or the browser is compromised (via malicious extension for example), an attacker could potentially grab the authorization grant and claim the data before you.
The access token, however, can only be claimed on a back channel. In addition to the authorization grant, the server also sends a secret key that only the client (server) knows. This is to let the resource server knows that this client did not steal the authorization grant. This secret key never goes to the front channel.
There exist an implicit flow which does not require an authorization grant. In this case, the client asks for the token directly. It is less secure for the reason cited above and should not be used if there is a way to use the back channel (it is mainly used by single page application with no back end).
As OAuth 2.0 was not designed to do authentication, some smart people added a little bit of whatever OAuth is missing for doing proper authentication use case. Out of that effort comes OpenID Connect.
OpenID Connect adds:
- ID token: represents the user or the information about the user
- UserInfo endpoint for getting more user information: used to retrieve user information
- Standard set of scopes
- Standardized implementation
The authorization flow is similar to OAuth 2.0’s; the only difference is as a
Scope we ask for
openid profile. Technically it’s both a OAuth 2.0 request and OpenID Connect request.
The ID token is A JSON web token (JWT) which is a standard way of encoding a bunch of information in a way that’s easy to transmit over the internet. It can be decoded to reveal a header, a payload (or claim) and a signature. The signature is a way to verify the payload has not been modified or forged before reception.