hi.
I was very surprised to hear Rhys's presentation on
draft-ietf-abfab-aaa-saml today, because Josh proposed to solve a
different problem than we've been working on over the last year.
Also, I do not feel very comfortable about the proposed solution.

As I understand it, in the case where we're not using SAML metadata, and
where we're relying on the AA trust fabric, we make the requirement that
the AAA entities correspond to the SAML entities.
So, we don't need to constrain the SAML naming because  the AAA entities
are making the assertions that the SAML names also correspond to the AAA
names.

In the case where we're trusting SAML metadata or policy, the issue is
that we need to avoid a cut&paste attack where a SAML message intended
for one party is intended for another party.  The issue we are trying to
solve is how to indicate in SAML messages or metadata that a particular
AAA endpoint was a valid destination for a binding or entity.

This issue is handled in the web profile by having the IDP redirect back
to the service provider.  We have more complexity because rather than
sending the message directly over HTTPS, the home AAA server sends the
message back over AAA.  we need to make sure that the home AAA server is
not man-in-the-middling the IDP--being a legitimate SP towards the IDP,
getting an assertion, and then using that assertion to give one of its
users access.

We've had a lot of discussion on this already, and I am fairly sure we
already rejected the approach of constraining the entity identifiers of
the SAML endpoints.

The approach proposed today is one that I prefer less well than
approaches Josh, Rhys and I have discussed offline.

_______________________________________________
abfab mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/abfab

Reply via email to