OAuth Confidential vs Public Client - What to choose if possible? - oauth-2.0

I understand the difference between Public and Confidential clients.
But what should be used if there is a choice?
Is Confidential Client preferable?

Well that depends on your specific use case. Since you haven't specified it clearly, the thumb rule is we should aim for higher security. This in turn means opting for Confidential client.
Please visit this link to understand when Confidential or Public client should be used depending upon the various use cases.
https://learn.microsoft.com/en-us/azure/active-directory/develop/msal-client-applications

Related

How does WebAuthn allow dependent web API's to access public key for decrypting credential without having to send the key?

I have familiarity with OAuth 2.0 / OpenID Connect but am new to WebAuthn. I am trying to understand how a scenario using those OAuth flows and connections would work using WebAuthn. I thought by mapping concepts from oauth to webauthn I would be able better understand the concepts.
I think similar to how in OAuth implicit grant flow a client may receive an id_token and access_token, in WebAuthn a client may receive a credential object from the Authenticator using navigator.credential.create.
The part I do not understand is how this credential can reliably be consumed by downstream services. In OAuth a client or server may send "access_tokens" and the receiving servers may request the public keys from the authorities to validate that it hasn't been tampered, is not expired, has correct audience, etc. This relies on the authorities having a publicly available /.well-known endpoint with the public keys.
However, I think because the keys are specific to the authenticator instead of a single shared public key it is not possible to have these be discoverable.
This is where I don't understand how credentials could be consumed by services. I thought the client would have to send the public key WITH the authenticator and client data but this is 3 pieces of information and awkward. Sending a single access_token seems actually cleaner.
I created a graphic to explain visually.
(It may have technical inaccuracies, but hopefully the larger point is made clearer)
https://excalidraw.com/#json=fIacaTAOUQ9GVgsrJMOPr,yYDVJsmuXos0GfX_Y4fLRQ
Here are the 3 questions embedded in the image:
What data does the client need to send to the server in order for the server to use the data? (Similar to sending access_token)
How would sever get the public key to decrypt data?
Which piece of data is appropriate / standardized to use as the stable user id?
As someone else mentioned - where there are a lot of commonalities between how WebAuthn and something like OpenID Connect work, they aren't really useful for understanding how WebAuthn works - they are better to explore after you understand WebAuthn.
A WebAuthn relying party does not have its own cryptographic keys or secrets or persistent configuration - it just has a relying party identifier, which is typically the web origin. The client (browser and/or platform) mediate between the relying party and authenticators, mostly protecting user privacy, consent, and providing phishing protection.
The relying party will create a new credential (e.g. key pair) with the authenticator of a user's choosing, be it a cell phone or a physical security key fob in their pocket. The response is the public key of a newly created key pair on the authenticator. That public key is saved against the user account by the RP.
In a future authentication, the authentication request results in a response signed by that public key. The private portion is never meant to leave the authenticator - at least not without cryptographic protections.
This does pair well with something like OpenID Connect. The registration is normally by web domain, which means that there could be a lot of manual registrations necessary (and potentially management, and recovery, and other IAM type activities) necessary. With OpenID Connect, you can centralize the authentication of several applications at a single point, and with it centralize all WebAuthn credential management.
I thought by mapping concepts from oauth to webauthn I would be able better understand the concepts.
This seems to be working against you - you're trying to pattern match WebAuthn onto a solution for a different kind of problem (access delegation). Overloaded terminology around "authentication" doesn't help, but the WebAuthn specification does make things a bit more clear when it describes what it means with "Relying Party":
Note: While the term Relying Party is also often used in other contexts (e.g., X.509 and OAuth), an entity acting as a Relying Party in one context is not necessarily a Relying Party in other contexts. In this specification, the term WebAuthn Relying Party is often shortened to be just Relying Party, and explicitly refers to a Relying Party in the WebAuthn context. Note that in any concrete instantiation a WebAuthn context may be embedded in a broader overall context, e.g., one based on OAuth.
Concretely: in your OAuth 2.0 diagram WebAuthn is used during step 2 "User enters credentials", the rest of it doesn't change. Passing the WebAuthn credentials to other servers is not how it's meant to be used, that's what OAuth is for.
To clarify one other question "how would sever get the public key to decrypt data?" - understand that WebAuthn doesn't encrypt anything. Some data (JS ArrayBuffers) from the authenticator response is typically base64 encoded, but otherwise the response is often passed to the server unaltered as JSON. The server uses the public key to verify the signature, this is either seen for the first time during registration, or retrieved from the database (belonging to a user account) during authentication.
EDIT: Added picture for a clearer understanding of how webauthn works, since it has nothing to do with OAuth2 / OpenID.
(source: https://passwordless.id/protocols/webauthn/1_introduction)
Interestingly enough, what I aim to do with Passwordless.ID is a free public identity provider using webauthn and compatible with OAuth2/OpenID.
Here is the demo of such a "Sign in" button working with OAuth2/OpenID:
https://passwordless-id.github.io/demo/
Please note that this is an early preview, still in development and somewhat lacking regarding the documentation. Nevertheless, it might be useful as working example.
That said, I sense some confusion in the question. So please let me emphasize that OAuth2 and WebAuthN are two completely distinct and unrelated protocols.
WebAuthN is a protocol to authenticate a user device. It is "Hey user, please sign me this challenge with your device to prove it's you"
OAuth2 is a protocol to authorize access to [part of] an API. It is "Hey API, please grant me permission to do this and that on behalf of the user".
OpenID builds on OAuth2 to basically say "Hey API, please allow me to read the user's standardized profile!".
WebauthN is not a replacement for OAuth2, they are 100% independent things. OAuth2 is to authorize (grant permissions) and is unrelated to how the user actually authenticates on the given system. It could be with username/password, it could be with SMS OTP ...and it could be with WebauthN.
There is a lot of good information in the other answers and comments which I encourage you to read. Although I thought it would be better to consolidate it in a single post which directly responds to the question from OP.
How does WebAuthN allow dependent web API's to access public key for decrypting credential without having to send the key?
There were problems with the question:
I used the word "decrypt" but this was wrong. The data sent is signed not encrypted and so key is not used to decrypted but verify the signature.
I was asking how a part of OAuth process can be done using WebAuthN however, this was misunderstanding. WebAuthN is not intended to solve this part of process so the question is less relevant and doesn't make sense to be answered directly.
Others have posted that WebAuthN can be used WITH OAuth so downstream systems can still receive JWTs and verify signatures as normal. How these two protocols are paired is a out of scope.
What data does the client need to send to the server in order for the server to use the data?
#Rafe answered: "table with user_id, credential_id, public_key and signature_counter"
See: https://www.w3.org/TR/webauthn-2/#authenticatormakecredential
How would server get the public key to decrypt data?
Again, decrypt is wrong word. Server is not decrypting only verifying signature
Also, the word server has multiple meanings based on context and it wasn't clarified in the question.
WebAuthN: For the server which acts as Relying Party in WebAuthN context, it will verify signature during authentication requests. However, the server in question was intended to mean the downstream APIs would not be part of WebAuthN.
OAuth: As explained by others, theses API servers could still be using OAuth and request public key from provider for verification and token contains necessary IDs and scopes/permissions. (Likely means able to re-use existing JWT middlewares)
Which piece of data is appropriate / standardized to use as the stable user id?
For WebAuthN the user object requires { id, name, displayName }. However, it intentionally does not try to standardize how the ID may propagated to downstream systems. That is up to developer.
See: https://www.w3.org/TR/webauthn-2/#dictdef-publickeycredentialuserentity
For OAuth
sub: REQUIRED. Subject Identifier. A locally unique and never reassigned identifier within the Issuer for the End-User
See: https://openid.net/specs/openid-connect-core-1_0.html#TokenResponse
Hopefully I didn't make too many technical inaccuracies. 😬

Is there any way to have confidential clients using javascript-adapter

I am very aware that you can't store your secret on a front end app, however, is there any way to work around this thus still having the benefits while using the javascript adapter.
I'm guessing using the JWT token option lead to the same issue.
I've read about using 2 different clients, one as a confidential admin and the other as a public client. All though I don't see how it is any better as the secret will still be held publicly, just in a different location.
Should I look deeper into this, are there any other workaround ?
Thanks
Solution : This won't turn your client into a confidential one, but you can add a layer of security by using PKCE.

PKCE vs DCR for OAuth2 and Mobile Applications

I’m developing my own OAuth2 + OpenID Connect implementation. I am a bit confused about how to handle OAuth flows for native (specifically, Mobile) clients. So far, I am seeing that I need to use an Authentication Code Flow. However, based on my research, there are some details that seem to contradict each other(at least based on my current understanding).
First, standard practice seems to say that mobile apps are not inherently private and, as such, standard flows that make use of a back channel should not be used. As a work around, the PKCE extension can be used (and utilize the built-in device browser as opposed to a web view so the tokens and sensitive information are less likely to be leaked).
However, under the Protocol’s Dynamic Client Registration specification, it is also mentioned that mobile apps should use this method of client registration to get a valid client ID and client secret... But, why would we do this when in an earlier section it was established that mobile applications were indeed public clients and couldn’t be trusted with confidential information like a client secret (which we are getting by using this DCR mechanism...
So, what am I not understanding? These two things seem to contradict one another. One claims mobile apps are public shouldn’t be trusted with a secret. Yet, in the recommended DCR mechanism, we assign them the secret we just established they can’t be trusted with.
Thanks.
A bit late, but hope it helps. So part of the OAuth2.0 protocol is two components, the client_id, and client secret. The client and server must agree on those two values outside the protocol i.e. before the protocol start. Usually, the process is as follows. The client communicates with the Authorization Server using an out-of-bound communication channel to get these values and be registered at the server. There is two way this client registration can happen, statically and dynamically. Statically mean the client_id and secret do change, i.e. the client gets them once when he registers with the server. Dynamic client registration refers to the process of registering a client_id every time the client wants to use to protocol, i.e. a client secret will be generated for him every time (also by an outbound communication).
Now, Why use dynamic registration?
Dynamic client registration is better at managing clients across replicated authorization servers., The original OAuth use cases revolved around single-location APIs, such as those from companies providing web services. These APIs require specialized clients to talk to them, and those clients will need to talk to only a single API provider. In these cases, it doesn’t seem unreasonable to expect client developers to put in the effort to register their client with the API, because there’s only one provider.
Does Dynamic Client registration offer any security advantages?
No, both are vulnerable if used with a JavaScript or a Native Mobile Client (JavaScript client can be inspected, and Mobile apps can be decompiled). Hence, both of them require PKCE as an extra layer of security.

OpenIddict local validation - multiple encryption keys for multiple Resourceservers

when i use local validation instead of introspection in my project with OpenIddict
usually the Authserver and the Resourceserver are sharing one symmetric encryptionkey. However when i use more than one Resourceserver i would like to use more than one symmetric encryptionkey (each Resourceserver should have it's own encryptionkey). Is there any way to achieve this?
Thanks for your help,
Nicolai
This is not currently supported.
It may be supported in a future version, but it has important design considerations: e.g the JWT compact format doesn't support embedding multiple content encryption keys, which prevents issuing encrypted JWTs to different recipients/resource servers.
If you don't want to share the secret with the resource servers (e.g. if you need to support 3rd party resource servers) you can use introspection.
But introspection has another drawback: I found no way to cache the introspection response in the openiddict validator. I don't think it is a good idea to have a roundtrip to the auth server on every API request. This could slow down high frequented API servers. Not sure if there is a simple solution for caching the response...

Should I use client_secret in a native, public downloadable application?

I've read a lot about the different flows (authorization code, implicit, hybrid and some extensions such as PKCE). Now I'm on the authorization code flow with PKCE.
PKCE ensures the initiator is the same user as the users who exchanges the authorization code for an access token. That is nice and OK.
When using this flow without a client_secret (which is recommended for SPA/Javscript applications) there is no warranty that the client is the known/original client. So, the 'consent' the user gave, is of no value. uhh?
I am working on a nativate client (a public downloadable binary). A secret cannot be considered confidential when baked in the binary, it can be decompiled for example.
Now I'm in dubio. What is better, bake the secret in the binary so that there is some extra layer of assurance the client is the known client or stop asking for 'consent' and give the same client_id to the whole world, only relying on the user-credentials.
Or is there something wrong with my story?
Very good question and made me realise a gap in my understanding. It is the role of the redirect uri to deal with this risk. In the web / https case the only hack that could work would be to edit the hosts file of the user. I'm the native case it is less perfect and your question is covered below. Generally our best bet is to follow recommendations / standards - but they have plenty of problems! https://web-in-security.blogspot.com/2017/01/pkce-what-cannot-be-protected.html?m=1
To others reading this case I've read a lot more.
Client impersonation is not easy fixable.
RFC8252 seems to be the most applicable article with recommendations for native apps - https://www.rfc-editor.org/rfc/rfc8252
"Claimed ‘https’ scheme" is mentioned as the best solution (IOS, Android and maybe UWP apps).
Since I'm working on a native Windows, non-UWP application I can't use this. As far as I can see the "Web Authentication Broker based on the app SID" is possible for my situation.
The other method is to accept the client as not known/identified and ask for 'consent' every time the client would access personal data.

Resources