I would differentiate re-authentication and re-keying. We
can debate
re-keying and the differences between re-authentication and
re-keying
later, but here let's talk about re-authentication.
Yoshi, it is understandable that you want to preserve the
word
authentication for full authentication. At a first glance,
I too
thought that is better. However, given the pass-through
model and
the use of that term in 4187, use of the word
re-authentication in
case of efficient re-authentication is apt.
Here is why: first, re-authentication, just as EAP full
authentication, results in an MSK delivery to the
pass-through
authenticator. The backed server uses a different process
for
authentication and a new rMSK derivation, but that's ok; as
long as
the parallels to the MSK derivation, and the use of EAP for
authentication are present, it is EAP (re-)authentication.
Besides,
4187 uses the terminology already: there fast
re-authentication is
defined as authentication "based on the keys derived
on the preceding
full authentication" as I quoted already. I put
together all this
information and convinced myself that the word
re-authentication
fits. Hope it's convincing to you too. Inventing another
term will
only create unnecessary confusion.
best,
Lakshminath
At 10:14 PM 9/24/2006, Narayanan, Vidya wrote:
>Hi Yoshi,
> > >
> >
> > I think that whether the signaling is
method-dependent or not
> > is not the fundamental difference between
method-based "fast"
> > re-authentication and method-independent
"efficient"
> > re-authentication. The fundamental difference
seems to be
> > whether the signaling results in re-keying of EAP
keying
> > material (MSK, EMSK,
> > TEKs) or not. I think any method-based
re-authentication
> > requires re-keying of EAP keying material while
> > method-independent "efficient"
> > re-authentication does not. Instead,
method-independent "efficient"
> > re-authentication would require re-keying of a
child key that
> > is derived from some intermediate key that is
further derived
> > from the EAP keying material (when it is used for
the serving
> > authenticator) or would require initial generation
of a child
> > key (not re-keying an existing child key) from the
> > intermediate key (when it is used for target
authenticators).
> >
>
>Your observation is mostly correct.
>
> > Also, it might be easier to understand to use the
word
> > "authentication" in the EAP keying
framework draft ONLY WHEN
> > we talk about any type of signaling that requires
EAP run
> > (e.g., EAP pre-"authentication" and
EAP re-"authentication").
> >
> > Having said that, it might be good to call
> > "method-independent efficient
re-authentication" something
> > else (e.g., intermediate keying). I just feel
confusing to
> > use the word "authentication" for it.
> >
>
>I don't quite agree with the above. Note that method
independent
>efficient re-authentication (as the method-specific
ones) actually does
>re-authenticate the peer by verifying proof of
possession of keying
>material generated from a previous EAP run.
>
> > BTW, I don't much like the terms
"fast" and "efficient" which
> > are ambigous in many cases.
> >
>
>I believe both method-specific and method-independent
re-authentication
>are in fact, re-authentication and re-keying. They
re-key different
>keying material, as you observe, but for the same
purpose. It is just
>that the adjective "fast" has history and
"efficient" was chosen to
>differentiate from that. I personally am okay with
efficient, since the
>goal is a single roundtrip protocol anyway - but, do you
have other
>suggestions in mind?
>
>Vidya
____________________________________________________________
_____
To unsubscribe or modify your subscription options, please
visit:
http:/
/lists.frascone.com/mailman/listinfo/eap
Arhives: http://lists.
frascone.com/pipermail/eap
|