On Sat, Feb 17, 2007 at 11:43:38PM -0800, Narayanan, Vidya wrote:
Yes, the problem of an authenticator providing different identities to
the peer and the server is the typical channel binding problem and can
be detected by simply doing a protected exchange of the identity between
the peer and server. When such a discrepancy is detected, then, keys
won't be handed out or if the identity is part of the key derivation,
then, it will result in a key mismatch anyway. Hence, there is no
problem there.
In my understanding, Dan's claim is that the server is unable to detect
that an authenticator is claiming an incorrect identity and by virtue of
that, if the authenticator claims the false identity to both the peer
and the server, a key will be provided to the authenticator and that
will match the key that the peer derives, even if the identity was part
of the key derivation. This is the problem that I have detailed in my
earlier email and I belive that can be resolved with the text I
proposed.
Yes, this is a problem that should not happen. If this happens, the
peer may send out data that is specific to ISP2 to the lying
authenticator (which is actually an ISP1 entity but claiming itself
to be an ISP2 entity) that is not supposed to see the ISP2-specific
data.
Going back to your proposed text:
"It is RECOMMENDED that the key transport protocol be able to detect
impersonation. When it is not feasible to guarantee that, every key
handed out from the server to an entity for a given peer MUST be
different from every other key handed out for a given peer."
I think that detection of impersonation is part of the "Authenticate
all parties" *requirement* (not a recommendation).
By the way, recent discussion on 3-party key distribution over HOKEY
mailing list made me think more about the following text in Section 5
(I don't know whether this is part of Dan's comment as I did not see
his original comment on this mailing list):
"
The authenticator is also a trusted party. The authenticator is
trusted not to distribute keying material provided by the AAA server
to any other parties. If the authenticator uses a key derivation
function to derive additional keying material, the authenticator is
trusted to distribute the derived keying material only to the
appropriate party that is known to the peer, and no other party.
When this approach is used, care must be taken to ensure that the
resulting key management system meets all of the principles in this
document, confirming that keys used to protect data are to be known
only by the peer and authenticator.
"
I understand that the authenticator is a trusted party for the EAP/AAA
server. On the other hand, I don't think we can say that it is a
trusted party for the peer before running EAP, unless the link between
the peer and the authenticator is already secured or the peer and the
authenticator have already a shared key to believe before running EAP.
Regards,
Yoshihiro Ohba
Regards,
Vidya
-----Original Message-----
From: Lakshminath Dondeti [mailto:ldondeti(_at_)qualcomm(_dot_)com]
Sent: Saturday, February 17, 2007 9:36 AM
To: Sam Hartman
Cc: Narayanan, Vidya; bernarda(_at_)microsoft(_dot_)com; Dan Harkins;
ietf(_at_)ietf(_dot_)org
Subject: Re: comments on draft-houseley-aaa-key-mgmt-07.txt
Sam,
The problem of an entity in the middle giving disparate
information to the peer and the server is in fact easier to
solve than the problem Vidya summarized. The disparate
information problem has been described in the EAP Keying
Framework document and elsewhere too.
To my understanding, we are beyond that point in the
discussion in HOKEY and considering the new case of the
entity in the middle lying to both sides and attempting to
get a key that another entity in the middle is supposed to get.
Let me put it this way, both issues are considered problems
to address/solve in this case.
regards,
Lakshminath
Sam Hartman wrote:
Vidya, I found the model you proposed didn't fit what Dan
was talking
about very well. In particular, Dan wants to focus on problems
resulting from the fact that the name of the authenticator used
between the peer and the authenticator may be different
than the name
of the authenticator used between the authenticator and the AAA
server. That distinction did not figure prominently enough in your
argument that I could tell whether you and Dan are talking
about the
same thing nor whether I could even tell if I agreed with you. I'd
recommend refocusing your model on this distinction; I
think once you
do we may well make significant progress on discussing a
long-standing
issue.
--Sam
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf