Authorization across origins

Hi,

I have a back-end. A type of self-hosted CMS. This CMS maintains tokens; obtained upon initial login or user creation, maintained and refreshed, and encrypted at rest.

To perform its operations, this CMS uses a M2M “main” token.

I have a Jamstack-type front-end on a different server, different domain. That front-end commonly uses an API, registered with Auth0, and uses passwordless, with RS256. For management-type operations, that front-end uses its own “main” client id and client secret.

But there are times when that front-end needs to tap data stored in the CMS.

I am wondering what the best strategy is here. Should I register the FE client id and client secret as additional keys with the CMS, so the CMS can “impersonate” the front-end?

Should I treat the front-end and CMS as a single app with shared keys? But they live on different domains.

The CMS needs to be able to validate the signature of requesting tokens coming from the front-end basically. It might even need to retrieve “userinfo”, and use custom claims to perform some authorization.

Pointers welcome. Thanks.

Hi @argo,

If I understand correctly, you should be using a Single-Page Applications with API strategy. With your FE app requesting access tokens and sending them with requests to your CMS/API. The CMS/API would validate the token and serve the data.

Is the front end requesting data from the CMS? Can you elaborate on why the CMS would be impersonating the FE? I am a little bit confused by this part.

Hi @dan.woda , thanks for your reply.

  • The front-end is requesting both data from the CMS, and from a separate API.
  • The CMS also accesses data from that API, independently. In addition to its own data.

It’s worth noting that that API is not fully fleshed-out on Auth0. We just use minimal custom claims for access control.

When a request comes from the FE, it is unauthenticated. After we verify and validate the token, we add a custom claim to match that Auth0 user to an internal ID from the CMS.

To generate the internal ID, we use a piece of unique data - ie. an email or similarly unique piece of info, which we pull from /userprofile.

At this point, the only token from which it seems safe to do this is the FE token. Hence, the CMS has to be able to perform operations on behalf of the front-end.

Thanks for expanding on that. In my mind, you would have the front end registered as a SPA app, the CMS registered as an M2M app and API, and the other backend registered as an API.

  • The user would authenticate in the FE
  • When the FE app needs data from the CMS or other API it would request an access token for either API (audience)
  • Respective access token would be sent with requests to the CMS or other API
  • If the CMS needs to make requests to the other API it would request a M2M token
  • M2M token would be sent with requests to the other API

I’m a little confused about this part:

How is the request from the FE unauthenticated while still having a token? Also, how are you adding custom claims to the token after the token has been issued from Auth0?

Hey @dan.woda

Thanks for taking the time.

Right now, this is mostly what we have. Except that the CMS is not currently registered as an API.

I am referring to CMS-internal authentication. The CMS has its own user management. We want to match that to incoming tokens.

That’s not what I meant. Custom claims are added via hooks/actions when the tokens are issued. Let me try to clarify.

  • A user logs in on the front-end. Using passwordless authentication.
  • They receive a token with some custom claims added.
  • They issue a request to the CMS for whatever data they need.
  • The CMS verifies the token signature.
  • If the CMS doesn’t have an internal user to match the one on the front-end, it wants some user details.

This is where I need to issue a call to userinfo, and the CMS, in my understanding, needs to act on behalf of the front-end.

At this point, the only guarantee I have on identity is that someone has sent a JWT that was, in all likelihood, issued by Auth0.

  • Once the CMS retrieves the userinfo, either this userinfo contains a claim that matches a user ID on the CMS…
  • Or the user/user id are created, and added via api call to the Auth0 user for this JWT (metadata).

That is very clarifying, thank you!

I think this set up makes sense how you have described. The access token can be exchanged for userinfo (alternatively, you should be able to inspect the token and grab the user ID/sub. You could also add more data as custom claims if there is a specific claim you need).

Did I miss any questions?

@dan.woda apologies for the very belated answer. No, I think this clarifies my use case.

My concern was: I could have used the m2m access from the CMS to get information about users, but this seemed to cross boundaries in a way I wasn’t comfortable with; besides, the PHP SDK somewhat prevents this.

JWKS verification requires the issuing app id and secret for a token, which means that I can only ever verify my front-end tokens from an app that knows the issuing client id and secret.

So: problem solved. The functionality works as intended. Thanks again.

1 Like

This topic was automatically closed 15 days after the last reply. New replies are no longer allowed.