|
List Info
Thread: issue 357: Channel Binding Definition
|
|
| issue 357: Channel Binding Definition |

|
2006-05-08 17:44:31 |
Hi Jari,
>
> Re: do all the three parties have to be involved in the
> binding? I believe that is essential for the definition
to say this.
>
Agreed.
> Re: the data that is subject to channel binding. I
think it
> is clear that not all data known by the parties needs
to be
> bound. Traditionally, we've only talked about security
or
> billing related information, such as the NAS identity
and
> type and cost of service. A large number of other
parameters
> may be known by some of the parties, but not relevant
from a
> verification perspective.
>
True. Actually, there does seem to be a line of thinking
where perhaps
channel binding capabilities can even be tied to the peer
verifying
advertised capabilities of the authenticator with the AAA
server. Of
course, it can be argued that it has nothing really to do
with channel
binding (and I'd agree) - but, goes to prove again that
this is an
opaque blob that may not be fully available at the peer.
> Re: do all the parties share the same data? I tend to
agree
> with Vidya here that its sufficient to for the verifier
to
> check that the data it gets is correct. But this issue
is
> also related to policies of what the parties want to be
> verified. If the server wants to verify FOO but the
client
> does not send it, we're not helping. This leads me to
think
> that the channel bindings, if we ever get them
deployed, are
> going to have use a well standardized set of parameters
or
> else we have a policy management problem in our hands.
But
> see below for a process comment.
>
> Process comment: Lets try to focus on the definition of
> channel bindings rather than their implementation in
this
> discussion. EAP keying framework does not define how to
> implement them, or whether to pick solution 1, 2, or 3.
>
Fully agree with this. Let's define it for the EAP keying
document and
continue to discuss solutions later for a separate document.
> Suggested text:
>
> "Channel Binding
>
> A secure mechanism for ensuring that a chosen set of
channel
> properties (such as endpoint identifiers) are agreed
upon by
> the EAP peer, authenticator and server."
>
> Vidya: note that the "chosen set" may be
different in
> different situations, depending on what information is
available.
>
Agree with the chosen set part. But, what is the
authenticator agreeing
to? Isn't it just an agreement about the authenticator
between the peer
and server?
Vidya
> --Jari
>
>
>
____________________________________________________________
_____
> To unsubscribe or modify your subscription options,
please visit:
> http:/
/lists.frascone.com/mailman/listinfo/eap
>
> Arhives: http://lists.
frascone.com/pipermail/eap
>
____________________________________________________________
_____
To unsubscribe or modify your subscription options, please
visit:
http:/
/lists.frascone.com/mailman/listinfo/eap
Arhives: http://lists.
frascone.com/pipermail/eap
|
|
| issue 357: Channel Binding Definition |

|
2006-05-09 10:46:56 |
Hi Vidya, Laksminath,
Vidya you wrote:
>>Suggested text:
>>
>>"Channel Binding
>>
>>A secure mechanism for ensuring that a chosen set of
channel
>>properties (such as endpoint identifiers) are agreed
upon by
>>the EAP peer, authenticator and server."
>>
>>Vidya: note that the "chosen set" may be
different in
>>different situations, depending on what information
is available.
>>
>>
>>
>
>Agree with the chosen set part. But, what is the
authenticator agreeing
>to? Isn't it just an agreement about the authenticator
between the peer
>and server?
>
>
And you Laksminath wrote:
> This is confusing to me. I was under the assumption
that the peer and
> the server should agree that they have the same view of
the
> authenticator (perhaps with the provision that the
server may have a
> superset of the information than the peer has). What
does the
> authenticator need to agree on? Does it matter?
Good questions. I'd be happy to restrict the definition
to peer and server agreeing that they have the same
view of the channel properties claimed by the authenticator.
(But part of the distinction may also be in the specific
implementation of the "agreement"; what we are
looking
for is that the values agree, without specifying who sends
the values and who verifies them.)
--Jari
____________________________________________________________
_____
To unsubscribe or modify your subscription options, please
visit:
http:/
/lists.frascone.com/mailman/listinfo/eap
Arhives: http://lists.
frascone.com/pipermail/eap
|
|
| issue 357: Channel Binding Definition |

|
2006-05-09 14:02:43 |
On Tue, May 09, 2006 at 01:46:56PM +0300, Jari Arkko wrote:
> And you Laksminath wrote:
>
> > This is confusing to me. I was under the
assumption that the peer and
> > the server should agree that they have the same
view of the
> > authenticator (perhaps with the provision that the
server may have a
> > superset of the information than the peer has).
What does the
> > authenticator need to agree on? Does it matter?
>
> Good questions. I'd be happy to restrict the
definition
> to peer and server agreeing that they have the same
> view of the channel properties claimed by the
authenticator.
If the definition is only for agreement between the peer and
server,
then it means we are allowing Channel Binding without
running a SAP.
I think involvement of the authenticator's agreement via
SAP is
needed. I mean the authenticator needs to agree that, the
authenticator, not someone else, has sent the information to
the peer.
Yoshihiro Ohba
____________________________________________________________
_____
To unsubscribe or modify your subscription options, please
visit:
http:/
/lists.frascone.com/mailman/listinfo/eap
Arhives: http://lists.
frascone.com/pipermail/eap
|
|
| issue 357: Channel Binding Definition |

|
2006-05-09 16:30:52 |
At 07:02 AM 5/9/2006, Yoshihiro Ohba wrote:
>On Tue, May 09, 2006 at 01:46:56PM +0300, Jari Arkko
wrote:
> > And you Laksminath wrote:
> >
> > > This is confusing to me. I was under the
assumption that the peer and
> > > the server should agree that they have the
same view of the
> > > authenticator (perhaps with the provision
that the server may have a
> > > superset of the information than the peer
has). What does the
> > > authenticator need to agree on? Does it
matter?
> >
> > Good questions. I'd be happy to restrict the
definition
> > to peer and server agreeing that they have the
same
> > view of the channel properties claimed by the
authenticator.
>
>If the definition is only for agreement between the peer
and server,
>then it means we are allowing Channel Binding without
running a SAP.
>I think involvement of the authenticator's agreement
via SAP is
>needed. I mean the authenticator needs to agree that,
the
>authenticator, not someone else, has sent the
information to the peer.
Hmmm, let's see. The peer and the server agree, let's say
through the
EAP method, that they are talking to the same authenticator
and the
server sends the MSK (or for that matter, as Joe seems to
imply, port
open indication) to *that* authenticator. So, I don't see
why the
authenticator needs any more verification.
Now without the SAP there are other problems, but they
don't seem
relevant to CB. Perhaps I am missing something (you seem
like you've
explored this area quite a bit, please explain. Thanks).
regards,
Lakshminath
>Yoshihiro Ohba
____________________________________________________________
_____
To unsubscribe or modify your subscription options, please
visit:
http:/
/lists.frascone.com/mailman/listinfo/eap
Arhives: http://lists.
frascone.com/pipermail/eap
|
|
| issue 357: Channel Binding Definition |

|
2006-05-09 19:00:37 |
On Tue, May 09, 2006 at 09:30:52AM -0700, Lakshminath
Dondeti wrote:
> At 07:02 AM 5/9/2006, Yoshihiro Ohba wrote:
> >On Tue, May 09, 2006 at 01:46:56PM +0300, Jari
Arkko wrote:
> >> And you Laksminath wrote:
> >>
> >> > This is confusing to me. I was under the
assumption that the peer and
> >> > the server should agree that they have
the same view of the
> >> > authenticator (perhaps with the provision
that the server may have a
> >> > superset of the information than the peer
has). What does the
> >> > authenticator need to agree on? Does it
matter?
> >>
> >> Good questions. I'd be happy to restrict the
definition
> >> to peer and server agreeing that they have the
same
> >> view of the channel properties claimed by the
authenticator.
> >
> >If the definition is only for agreement between the
peer and server,
> >then it means we are allowing Channel Binding
without running a SAP.
> >I think involvement of the authenticator's
agreement via SAP is
> >needed. I mean the authenticator needs to agree
that, the
> >authenticator, not someone else, has sent the
information to the peer.
>
> Hmmm, let's see. The peer and the server agree, let's
say through the
> EAP method, that they are talking to the same
authenticator and the
> server sends the MSK (or for that matter, as Joe seems
to imply, port
> open indication) to *that* authenticator. So, I don't
see why the
> authenticator needs any more verification.
The problem here is that the authenticator that advertised
the
information may not be the one that receives the MSK.
>
> Now without the SAP there are other problems, but they
don't seem
> relevant to CB. Perhaps I am missing something (you
seem like you've
> explored this area quite a bit, please explain.
Thanks).
Please see an example below.
[Peer]---[AA]---[LA]---[Server]
| | |
| EAP method | |
|<------------------>|
| |MSK |
| |<---|
An attacker authenticator (AA) is acting as an authenticator
for the
peer and and as a peer for the legitimate authenticator
(LA), passing
through EAP messages between the peer and LA. Both AA and
LA
advertise the same LA's information XYZ. The peer learns
XYZ from AA.
The server have the XYZ for LA. Server sends MSK to LA
after creating
binding between XYZ and MSK by some means (i.e., key mixing
or data
transfer).
If we stop here, the peer will think that AA has the MSK,
which is not
true. This looks like an open binding which might open up
some
vulnerability to MiTM attack. An additional verification,
SAP, is
needed to close the binding by verifying proof of possession
of the
MSK between the peer and authenticator.
Regards,
Yoshihiro Ohba
____________________________________________________________
_____
To unsubscribe or modify your subscription options, please
visit:
http:/
/lists.frascone.com/mailman/listinfo/eap
Arhives: http://lists.
frascone.com/pipermail/eap
|
|
| issue 357: Channel Binding Definition |

|
2006-05-09 20:39:28 |
At 12:00 PM 5/9/2006, Yoshihiro Ohba wrote:
>On Tue, May 09, 2006 at 09:30:52AM -0700, Lakshminath
Dondeti wrote:
> > At 07:02 AM 5/9/2006, Yoshihiro Ohba wrote:
> > >On Tue, May 09, 2006 at 01:46:56PM +0300, Jari
Arkko wrote:
> > >> And you Laksminath wrote:
> > >>
> > >> > This is confusing to me. I was
under the assumption that the peer and
> > >> > the server should agree that they
have the same view of the
> > >> > authenticator (perhaps with the
provision that the server may have a
> > >> > superset of the information than the
peer has). What does the
> > >> > authenticator need to agree on?
Does it matter?
> > >>
> > >> Good questions. I'd be happy to restrict
the definition
> > >> to peer and server agreeing that they
have the same
> > >> view of the channel properties claimed by
the authenticator.
> > >
> > >If the definition is only for agreement
between the peer and server,
> > >then it means we are allowing Channel Binding
without running a SAP.
> > >I think involvement of the authenticator's
agreement via SAP is
> > >needed. I mean the authenticator needs to
agree that, the
> > >authenticator, not someone else, has sent the
information to the peer.
> >
> > Hmmm, let's see. The peer and the server agree,
let's say through the
> > EAP method, that they are talking to the same
authenticator and the
> > server sends the MSK (or for that matter, as Joe
seems to imply, port
> > open indication) to *that* authenticator. So, I
don't see why the
> > authenticator needs any more verification.
>
>The problem here is that the authenticator that
advertised the
>information may not be the one that receives the MSK.
>
> >
> > Now without the SAP there are other problems, but
they don't seem
> > relevant to CB. Perhaps I am missing something
(you seem like you've
> > explored this area quite a bit, please explain.
Thanks).
>
>Please see an example below.
>
>[Peer]---[AA]---[LA]---[Server]
> | | |
> | EAP method | |
> |<------------------>|
> | |MSK |
> | |<---|
>
>An attacker authenticator (AA) is acting as an
authenticator for the
>peer and and as a peer for the legitimate authenticator
(LA), passing
>through EAP messages between the peer and LA. Both AA
and LA
>advertise the same LA's information XYZ. The peer
learns XYZ from AA.
>The server have the XYZ for LA. Server sends MSK to LA
after creating
>binding between XYZ and MSK by some means (i.e., key
mixing or data
>transfer).
>
>If we stop here, the peer will think that AA has the
MSK, which is not
>true. This looks like an open binding which might open
up some
>vulnerability to MiTM attack. An additional
verification, SAP, is
>needed to close the binding by verifying proof of
possession of the
>MSK between the peer and authenticator.
Yoshi,
You've just made a case for key generating EAP methods and
the need
for a SAP. There is no dispute there. But, my earlier
point about
without SAP, there are other problems is still true, no? CB
is not
the only issue there.
Lakshminath
>Regards,
>Yoshihiro Ohba
____________________________________________________________
_____
To unsubscribe or modify your subscription options, please
visit:
http:/
/lists.frascone.com/mailman/listinfo/eap
Arhives: http://lists.
frascone.com/pipermail/eap
|
|
| issue 357: Channel Binding Definition |

|
2006-05-09 21:14:35 |
Lakshminath,
On Tue, May 09, 2006 at 01:39:28PM -0700, Lakshminath
Dondeti wrote:
> >>
> >> Now without the SAP there are other problems,
but they don't seem
> >> relevant to CB. Perhaps I am missing
something (you seem like you've
> >> explored this area quite a bit, please
explain. Thanks).
> >
> >Please see an example below.
> >
> >[Peer]---[AA]---[LA]---[Server]
> > | | |
> > | EAP method | |
> > |<------------------>|
> > | |MSK |
> > | |<---|
> >
> >An attacker authenticator (AA) is acting as an
authenticator for the
> >peer and and as a peer for the legitimate
authenticator (LA), passing
> >through EAP messages between the peer and LA. Both
AA and LA
> >advertise the same LA's information XYZ. The peer
learns XYZ from AA.
> >The server have the XYZ for LA. Server sends MSK
to LA after creating
> >binding between XYZ and MSK by some means (i.e.,
key mixing or data
> >transfer).
> >
> >If we stop here, the peer will think that AA has
the MSK, which is not
> >true. This looks like an open binding which might
open up some
> >vulnerability to MiTM attack. An additional
verification, SAP, is
> >needed to close the binding by verifying proof of
possession of the
> >MSK between the peer and authenticator.
>
> Yoshi,
>
> You've just made a case for key generating EAP methods
and the need
> for a SAP. There is no dispute there. But, my earlier
point about
> without SAP, there are other problems is still true,
no? CB is not
> the only issue there.
I agree that CB is not the only issue here, but CB is *an*
issue. The
SAP for CB does not necessarily generate a TSK. In this
case, the
role of the SAP is just to close the binding. However, if
you do CB
but don't generate TSK and use it for per-packet ciphering,
then you
will surely have other issues.
Yoshihiro Ohba
____________________________________________________________
_____
To unsubscribe or modify your subscription options, please
visit:
http:/
/lists.frascone.com/mailman/listinfo/eap
Arhives: http://lists.
frascone.com/pipermail/eap
|
|
[1-7]
|
|