Replay attack prevention through AS contributed nonce

[tplooker]

Currently the specification does not define a mechanism for an AS contributed nonce to feature in the client attestation PoP meaning replay attack detection by authorisation servers requires tracking the JTI within the time window set by the PoP's iat claim. We should elaborate on an AS contributed nonce for the specification.

[paulbastian]

What about using the authorization code or pre auth code as suggested in our demo?

[tplooker]

I think this is an option, however the nonce for client authentication becomes dependent on the grant type which creates a bit of an awkward coupling when an application wants to say use this mode of client authentication for grant type other than pre auth code or authorisation code.

[paulbastian]

How many implemantations out there actually use different grant types than auth-code?
Given that this serves our purposes and imo everybody is using auth-code, I think that we should RECOMMEND that scenario but I#m ok with showing up alternatives as you suggested.
@tlodderstedt What's your opinion on this?

[tlodderstedt]

I think the flow can be used with and without nonce.

If it is used without nonce, jti and short expiration must be used to limit the risk.
If a nonce is used, I think using the codes is a pragmatic option. Alternatively, the AS could provide a dedicated nonce with the credential offer or authorization response. It could also provide the nonce with a token error response.

[paulbastian]

discussion on 23.05 is leaning towards DpoP-style error providing nonce

[tlodderstedt]

can we please bring this issue forward, it is very important?

discussion on 23.05 is leaning towards DpoP-style error providing nonce

I see some advantages in that approach as it would allow the AS to request nonce use for all endpoints (including PAR).

[tlodderstedt]

on that topic, I would suggest such an AS provided nonce should be good up until the AS says otherwise. That would allow a client to use the same nonce for PAR and token endpoint.

[tplooker]

Update provided in #39

[paulbastian]

PR #49

[paulbastian]

PR #49