W3C

Joint Task Force of WebAuthn and Web Payments

27 Nov 2019

Agenda

Attendees

Present
Ian Jacobs (W3C), Gerhard Oosthuizen (Entersekt), Nicolas Ancot (Canton), Lauren Helt (American Express), John Fontana (Yubico), Benjamin Tidor (Stripe), Tony Nadalin (Microsoft), Ken Buchanan (Google), Nick Telford-Reed, Tomasz Blachowicz (Mastercard), Jeff Hodges (Google), John_Bradley (Yubico), Jonathan Grossar (Mastercard)
Chair
Ian
Scribe
Ian

Contents


<jfontana> Ian, I need the WebEx URL

Deliverables

IJ: Which deliverables would be useful?

Nicolas: +1 to education materials.

Lauren: +1

John: +1 to use cases

Jonathan: +1 to use cases documentation

<jeffh> +1 to use cases

<kenrb> +1 from me also

IJ: Any templates people like for use case documentation?

IJ: Suggest consumer journey.

Jonathan: +1 to consumer journeys for enrollment, then accessing credentials, then transaction

Some use cases

IJ: We started with the "from an iframe" use case last week.

Jonathan: ... I am a consumer, I am on a merchant site, I want to register authenticator. How can I be sure I can register my authenticator?
... what does the merchant need to do in advance?
... for Payment, I need to access my credentials but also want to authenticate to my bank during the transaction.

USE case: Share authentication results so bank is happy

IJ: To whom did I first authenticate to?

Jonathan: Generically, I could auth to a merchant, to the owner of a payment method, to a payment handler.
... we need to look into these different flows.
... as the user I don't want to have to authenticate too much
... how can I leverage FIDO once and share results.

IJ: What is the WebAuthn description of this?

JeffH: We don't have that at this time. the term we've been using to label this is "delegation."
... in the WebAuthn context, it is presently unsolved and unaddressed in the specs.
... we are discussing ways to possibly address it in the next rev of the spec.

Jonathan: That would be interesting to discuss in this group.
... there might be different options, e.g., if credentials are generated and owned by merchant, or whether owned and generated by the issuing bank.

IJ: Is it useful to be specific: first merchant, then issuing bank?
... I ask this because it might relate to trust boundaries

Gerhard: I liked Jonathan's description. We might want to list separate use cases:

a) Merchant wants to leverage issuer credentials.

b) Merchant wants to pass on to issuer its credentials as proof that it's done webauthn.

scribe: I think that both use cases are useful.

Jonathan: +1

JeffH: From the user's perspective, what you are talking about is a single event.

Gerhard: Correct.

Jonathan: I agree both (a) and (b) are interesting. For (b), I think there might be some suboptions:

b.1) Issuer trusts the merchant's authentication (and some high level results are provided)

b.2) Issuer does not necessarily want to trust the merchant, and basically it amounts to data sharing between FIDO servers.

JeffH: When you say "the issuer does not step up" you mean, the issuer does not perform an explicit auth event that interrupts the user's task.

Jonathan: Yes

JeffH: I hear that you don't want to interrupt the user's task. The goal I hear is to not interrupt the user flow, not do a full page navigation, etc.

Jonathan: We need to deep dive into how to give issuer confidence.

JeffH: Threat model is important here.
... I'm not sure that having the issuer share credentials with the merchant is workable given the flow that the payment folks are wishing to work with.

[IJ mentions trust tokens]

JeffH: Trust tokens are imaginary at this time.

Ken: Another flow was proposed by some colleagues at Google - the user agent authenticates to the issuer, but initiated by the merchant.

Gerhard: Some of that is already done in 3DS. The challenge is that the merchant also wants to know a little bit about the user.
... the merchant wants to prevent illegal transfer, or to protect the merchant's assets
... so there are multiple situations where the merchant wants to provide context to the issuer.
... the issuer gets to make the choice of authenticating the user.
... it's the second authentication that could result in abandonment.
... I am thinking more about someone going to the banking site and installing an authentication payment handler, and the merchant can pass on some information to the issuer, but the issuer also gets a sort of pre-delegation for specific auth events that is passed along with payment info.
... and issuer almost has proof that the issuer was present up front.

Jonathan: We need to follow both routes.

IJ: Any volunteers?

JeffH: We (Google) are working on coming up with various approaches to this overall delegation problem. We will have stuff written down to inject into this discussion. Not sure when.
... maybe rough by the next call or after the holidays.

JeffH: At a high level we need to define the use cases from the consumer's perspective, merchant's perspective, and issuer's perspective. Understand the requirements. And analyze different solutions in the context of threat models.
... and how they map to the requirements, such that the use case is satisfied from those perspectives.

Next call

11 December

Summary of Action Items

Summary of Resolutions

[End of minutes]

Minutes manually created (not a transcript), formatted by David Booth's scribe.perl version 1.154 (CVS log)
$Date: 2019/11/27 17:08:10 $