Re: [Tcpcrypt] Draft charter text

[John-Mark Gurney <[e-mail]>]

Joe Touch wrote this message on Sun, Apr 13, 2014 at 11:10 -0700:
> On 4/13/2014 10:54 AM, Erik Nygren wrote:
> >On Sun, Apr 13, 2014 at 1:41 PM, Joe Touch <[e-mail]
> ><mailto:[e-mail]>> wrote:
> >
> >
> >    You need to MAC the IP dest, dest port, and all signal bits in the
> >    IP header (sequence, flags, etc.). This will still make it through a
> >    NAT, but will break a 'rewriting proxy' - but that's the kind of
> >    attack that TCP crypto ought to protect against, because it's
> >    indistinguishable from a MITM attack (it *is* a MITM).
> >
> >Unfortunately, applying a MAC on the IP dest will break NAT64 which is a
> >valid and increasingly critial use-case.
> 
> If you avoid MACing any address or port info (except perhaps dest port), 
> then all you have left to identify your connection is the ISN pair, and 
> that seems a bit dangerous.

How/Why is that dangerous?

If it is used w/ an authentication layer that authenticates that the
session hash is the same on both sides, how is this dangerous?  The
authentication layer has stated that the session key is known only to
the end points, and not to any MITM.

MAC'ing the IP/port info will prevent 90% of the internet from using
it was most end users are behind NATs...  Be it a corrporate nat, or
the nat their ISP provided so that they are only allocated one IP...

-- 
  John-Mark Gurney				Voice: [phone]

     "All that I will do, has been done, All that I have, has not."