Hi Sam:

Thanks for these clarifications. My comments inline again.

El 06/09/2011, a las 17:38, Sam Hartman escribió:

>>>>>> "Rafa" == Rafa Marin Lopez <[email protected]> writes:
> 
>    Rafa> Hi Sam: Please see my comments/questions inline.
> 
>    Rafa> El 06/09/2011, a las 15:52, Sam Hartman escribió:
> 
>>> 
>>> Hi.  At the mic in Quebec I raised the following issue about GSS
>>> preauth and ABFAB.
>>> 
>>> I want to be able to use GSS preauth between the RP and the KDC
>>> even if the initiator knows nothing about it.
> 
>    Rafa> Could you elaborate a little bit about the motivation related
>    Rafa> to this and what is the associated issue?
> 
> Sure.
> The idea is that you have some resource domain of servers.
> You'd like to use GSS-EAP (ABFAB) to access them.

> However  you want a central place to define policy, perform
> authorizations, etc.
> You've decided that Kerberos works well for that.
> All authentications into the domain must go through the KDC. Only the
> KDC can create authorizations.
> 
> You want the initiator to be able to take advantage of a TGT for fast
> reauthentication within a domain if the initiator understands how.

May we assume that TGT will be involved in a Kerberos exchange later on?. I 
mean I think that TGT will have to be provided to the initiator somehow ( 
within GSS-EAP exchange? )

I assume that initiator will have some Kerberos source code implemented to 
handle the TGT and to request service tickets. Otherwise, having a TGT is 
useless as you mention.

Thus I would assume that initiator has Kerberos source code implemented. is it 
so unreal that initiator would have a kerberos client?.  

Another question would be how the TGT will be used by the initiator and how 
everything will operate... where the keys will be distributed and how etc... 


> However, you want to work with any initiator. So, depending on
> gss-preauth-specific changes to the ininitiator isn't acceptable.

Mmmmm this seems to contradict your sentence "You'd like to use GSS-EAP (ABFAB) 
to access them." no?. At least it is assumed that the initiator has to 
implement GSS-EAP. So "any" initiator would not work. 

Also, it would be interesting to know why it may not be acceptable. After all, 
we are doing changes to include new technology :).

> In
> particular you don't want to have to bypass the KDC for old initiators.
> Services are not trusted to make authorization decisions.


Well, if they are old initiators most probably they will implement legacy 
mechanisms for authenticating services (not even GSS-EAP except, perhaps, 
Kerberos GSS-API mechanism, for example). On the other hand, services will need 
to be updated to support this (so no support for old services).

Apart from this interesting scenario you mention, I would also like to suggest 
to analyze another possible case for fast re-authentication, when the initiator 
has to go through the RP. The idea is simple if we can assume GSS-API 
pre-authentication mechanism in the initiator and IAKERB. As evident, GSS-EAP 
would be a potential mechanism to perform the GSS-API pre-authentication as it 
has been already discussed. Then the initiator will have a TGT and can request 
service tickets.

Initiator (GSS-API pre-authentication mech + IAKERB) --- RP (IAKERB proxy) --- 
KDC (Acceptor)

Here, KDC is also the central point of authentication and authorization. 
Certainly the assumptions/requirements differ a little bit in both cases but 
they seem interesting.

In any case, I believe it would be good to have a clearer picture of the 
general goal (e.g. to achieve fast re-authentication in GSS-EAP or so...). Then 
define requirements for a potential solution and needs.

Best regards.


> 
> 
> 
>>> 
>>> That is: gss-eap initiator <-> RP FAST(rp armor key,
>>> gss-preauth(gss-eap)) RP <->KDC When that happens:
> 
>    Rafa> Before going into detail I would like to understand this flow
>    Rafa> (at first sight, it vaguely reminded me the use of IAKERB
>    Rafa> where the RP is IAKERB proxy. But it seems you are pursuing a
>    Rafa> different thing). It seems that the RP is acting as Kerberos
>    Rafa> client that uses a GSS-API based pre-authentication mechanism
>    Rafa> and the content of the gss-api token contains the gss-eap
>    Rafa> token provided by the user. Is that right?
> 
> Your understanding is correct.
> 
> --Sam

-------------------------------------------------------
Rafael Marin Lopez, PhD
Dept. Information and Communications Engineering (DIIC)
Faculty of Computer Science-University of Murcia
30100 Murcia - Spain
Telf: +34868888501 Fax: +34868884151 e-mail: [email protected]
-------------------------------------------------------




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

Reply via email to