proxyAuth addon

[A Former User]

@girish Wow I totally didn't realize there were docs for it. Sorry for bugging you!

[girish]

@atrilahiji I just recently pushed it

[saikarthik]

@girish is there a way to get the username/email from within the app?

[nebulon]

@saikarthik currently not, I guess the only option would be to add the username/email as a header in the requests?

[jimcavoli]

@nebulon That would seem a sensible approach. Similar to other gateway authentication solutions I've seen. Definitely would need to restrict trust of those headers either in app or sever configuration though to prevent escalation/impersonation/ato attacks

[nebulon]

@jimcavoli is there any risk or impersonation angle, if the reverse proxy always explicitly overwrites that header?

[jimcavoli]

@nebulon Yes, an always-overwrite would mitigate as well, as long as the edges get tested well, might be the easier solution

[saikarthik]

@nebulon @girish is this something that can be added to cloudron? passing logged in username/email ID to apps through the header? Any comments/issues?

[girish]

@saikarthik yup, can surely be added. probably next release.

[jimcavoli]

Related: while re-working the n8n packaging, I happened upon what would probably be reasonably common, where there are selected sub-paths of / which should not be authenticated - example being we want / to require auth, but not /webhook/* paths. It's at least non-obvious if not unsupported by the current docs on how to do this with proxyAuth

[girish]

@jimcavoli Indeed, that's not something I designed for. How complicated can these rules get ? Atleast, https://docs.n8n.io/reference/security.html does not seems to have any more information. Or should I just add a publicPath property (singular) and that's enough ? I like to under-design these things and extend them as use cases come.

[mehdi]

@girish I think the best would be to have the path in proxyAuth be an array, where given paths can be either positive or negative. It's the way things like .gitignore work.

For example, in this case, it would be:

{
  "proxyAuth": [
    "/",
    "!/webbooks/"
  ]
}

[thetomester13]

@mehdi I like this solution and its flexibility. It could also be backwards compatible with the currently version - if no paths are specified, everything is auth'ed.

[jimcavoli]

Agree on the default behavior - I imagine it's unlikely that anything more specific than path-level exceptions are unlikely. Perhaps as an extension to the solution that @mehdi suggests, we could extend the existing format of:

{
  "proxyAuth": {
    "path": "/admin" 
  }
}

To take exceptions:

{
  "proxyAuth": {
    "path": "/admin" ,
    "exclude": [
      "/webhook",
      "/
    ]
  }
}

Or with probably over-the-top features, make everything a map of path and exception(s):

{
  "proxyAuth": {
    "paths": {
      "/" : [
        "/webhook",
        "/public"
      ],
      "/admin": []
    }
  }
}

Honestly, I appreciate the minimal-first approach, and I think the middle option of adding a (understood to be auto-wildcarded) array of exclusions is the easier next step. I can't imagine anything that would need the super-complex variant would be something that would or should rely on such a mechanism to secure it.

[nj]

@girish I don't see the 2FA code prompt on the login page of Simple Torrent. Am I missing something?

Some benefits of having this on the platform side (as opposed in the app are):

• 2FA login

[mehdi]

@nj I don't think this is implemented either:

• Session management in the user's profile page. i.e can logout from apps etc

I think @girish just meant that it would be possible to implement this in the future, not that it would be in the first version of proxyAuth.

[girish]

@nj I have logged it here - https://git.cloudron.io/cloudron/box/-/issues/748 . As @mehdi said, it wasn't implemented as part of the first iteration of proxyAuth.

[girish]

@jimcavoli Shall I go with path: "!/webhooks" for now? Will this be enough for n8n ?

[jimcavoli]

@girish Yeah, that would be enough for n8n I think, though if we're going to go that route, I think making paths an array of either path(s) and/or ! paths makes the most sense in general (at least somehow providing for the option of multiple excluded paths)

[girish]

@jimcavoli won't having it plural cause some confusing semantics if you mix ! and no ! paths ? Let me think