Is nonce required with JARM?

[Dima Postnikov]

Comment from Nat from a JARM thread:

This conversation also made me became aware of the subtle difference in the security posture. 

It may well be a bug of FAPI 1.0 Advanced. 

In the case of hybrid flow, `nonce` is required while in the case of JARM, it is not. 

So the binding characteristics between the authorization request and token response is a bit different between them, probably a bit weaker in the case of reponse_mode=jwt (JARM). 

‌

[Joseph Heenan]

I believe that nonce is required in FAPI1Adv when using JARM with scope=openid - as per https://bitbucket.org/openid/fapi/src/master/FAPI_1.0/openid-financial-api-part-1-1_0.md

If the client requests the openid scope, the authorization server

1. shall require the nonce parameter defined in Section 3.1.2.1 of OIDC in the authentication request.

(and if you’re not using scope=openid, nonce is not required, because there’s nowhere to return a nonce)

[Joseph Heenan]

(in the “not using scope=openid” case, state is required instead)

[Brian Campbell]

I came here to say what @Joseph Heenan said but will pedantically link to the published spec:

https://openid.net/specs/openid-financial-api-part-1-1_0.html#client-requesting-openid-scope

[Nat Sakimura]

And the design assumption was that it does, as it always had “openid” in the scope in the case without JARM so that there always is nonce and all the messages in one protocol execution would be threaded by the nonce. The introduction of JARM was supposed to provide only another mode and was not intended to produce a change in the security posture.

[Dave Tonge]

Will deal with in this ticket: https://bitbucket.org/openid/fapi/issues/571/document-differences-between-fapi1advanced