[OAUTH-WG] Proof of Possession Tokens: Next Steps

[Hannes Tschofenig <[e-mail]>]

Hi all,

I wanted to drop a high level message about possible next steps for the
PoP work.

As you have seen from my status update, see
http://www.ietf.org/mail-archive/web/oauth/current/msg15327.html, the
PoP architecture document was already in IESG processing but I have had
asked Kathleen to delay the publication given that we ran into scoping
issues, as discussed on the list. See
http://www.ietf.org/mail-archive/web/oauth/current/msg15177.html

The change of scope related to desire to not just binding a key to the
access token but also to other parts of the OAuth system to avoid cases
where an attacker can just obtain attack other parts of the system
instead (for example, by obtaining an bearer-based refresh token to then
obtain a new PoP access token).

The recently discovered security problems tell us that we need to
simplify our solutions a bit as well to ensure that we get the security
analysed properly. More options means more time to analyse all the
different options.

What does this mean to simplify when I talk about expanding the scope in
the earlier paragraph?

I am suggesting to

* to consider focusing on a public key-based only solution for the
web/smart phone app space. (The ACE working group will have to develop a
symmetric key-based version on their own, if desired.)

* to extend the support of PoP token functionality throughout the entire
solution. This means that we have to include support for a asymmetric
version of PKCE into account (which had been discussed in the group
already earlier already).

* to define at least a TLS-based security security solution for the
communication between the client and the resource server.

* to rethink the work on the application layer security solution. The
HTTP signing draft, which defines the application layer security
solution for use between the client and the resource server, has expired
and we will have to find new authors. I believe we got stuck a bit.
Luckily new persons came along and volunteered to help, namely Fredrik
Ljunggren and Jakob Schlyter. Nevertheless, the group will have to judge
whether a newly developed application layer security solution is
promising. My impression is that it is a very difficult to come up with
a solution that satisfies the security requirements and, at the same
time, also takes the deployment status of proxies and other middleware
into account.

* to make a decision about other extensions. Nat and Kepeng submitted
the Sender Constrained JWT for OAuth2 2.0 document, see
https://tools.ietf.org/html/draft-sakimura-oauth-rjwtprof-06
We asked the working group for feedback during IETF #93 and we couldn't
get enough feedback at that time. Please give us feedback whether you
are interested in exploring that solution direction as part of this
process. Today, we don't have enough indication of interest for working
on that solution direction.

Before making any changes to the PoP document set we would like to hear
your thoughts.

Ciao
Hannes