Re: [OAUTH-WG] Dynamic Scopes

[George Fletcher <[e-mail]>]

Thanks Torsten.

I think I have a solution :) Just to make sure I have the flow correct...

Assumption: Using a mobile client

1. User (using their mobile client) attempts to sign a document with the 
insurance company
2. Insurance company redirects the user to their Bank asking for 
identity proof, and signing of specific documents
3. User interacts with Bank to get authorization for the specific 
transaction
4. Mobile client submits request to insurance company using token that 
is specific to the user, document etc.

This is effectively the UMA 2.0 flow [1]

1. Mobile client attempts to invoke resource at the insurance company
2. Insurance company registers the request with UMA AS (the bank in this 
case) and gets a permissions ticket
3. Insurance company instructs mobile client to contact the bank
4. Mobile client contacts the bank specifying the permissions ticket
5. User meets banks requirements for the specific transaction (claims 
interaction)
6. Bank issues mobile client the RPT (token)
7. Mobile client invokes resource at insurance company with RPT

Note that the insurance company can specify the necessary bits that need 
to be in the token when it interacts with the Bank (as the UMA AS). 
[There might be some profiling required here]

I think it's worth exploring whether UMA will solve this use case.

Thanks,
George

[1] https://docs.kantarainitiative.org/uma/wg/oauth-uma-grant-2.0-08.html

On 6/23/18 3:43 AM, Torsten Lodderstedt wrote:
>
>> Am 22.06.2018 um 23:08 schrieb George Fletcher <[e-mail]>:
>>
>> I would think that the scope issued to the refresh_token could represent the category or class of authorizations the refresh_token should be able to perform. For example, the kind of transactions that can be bound to access tokens. The scope issued into the access_token could be one of the "parameterized" ones. But maybe I'm not fully understanding the use case :)
> Let me try to explain ;-)
>
> The client is an issuance company wanting the customer to electronically sign a new contract (legally binding!). Signing in the end means to send a request containing the hash of the document to an API. The API will respond with an CM/S Object containing signature, certificate etc that the client will embedded in the contract document (typical PDF).
>
> We want the user to authorize the signing request using their bank as IDP/AS. Therefore the client sends the OAuth authorization request to the AS. The actual signing request needs to be bound to client, user AND hash (document) in order to prevent fraud. Regulation (eIDAS) requires to always demonstrate the sole control of the user over the whole process. The AS therefore binds (scopes) the access token to exactly this single document/signing request. If the client wants the user to sign another document, it needs to got through the whole process again.
>
> One could think about a general signing permission represented by a refresh token, but not in the high assurance level cases I‘m looking into.
>
> Hope that helps,
> Torsten.
>
>