Re: [Spud] [Privsec-program] Detecting and Defeating TCP/IP Hypercookie Attacks

[Tom Herbert <[e-mail]>]

On Sun, Jul 31, 2016 at 11:34 PM, Eliot Lear <[e-mail]> wrote:
>
>
> On 7/30/16 9:59 PM, Tom Herbert wrote:
>>
>> No, it's not. In fact, exposing flow start/stop information is a good
>> example of something that facilitates intrusiveness by middleboxes at
>> the transport layer.
>>
>> The purpose of exposing this information is to allow network devices
>> to track connections, but connection tracking in the network is
>> fundamentally flawed since there is no requirement that all packets of
>> a connection go any single network device (i.e. the Internet is packet
>> switched not circuit switched).
>
> Except that I will bet you that over 99.999% of connections do, and that
> this is particularly sufficient for home routers.
>
Eliot,

If that number it is correct it is only because home routers have
ossified the Internet in that regard, not because the standard was
ever changed to require it. Desktops sitting behind home routers is no
longer a sufficient model for the Internet; mobile devices are
currently predominant and the Internet needs to adapt accordingly.

Consider that mobile devices are multihomed having at least two
network connections. We want the ability to seamlessly switch between
networks (say from wifi to mobile) or between mobile networks as we
drive down the road. Performing 3WHS is very expensive on mobile
(literally for some of our users), so we need connections to survive
across these path changes. If we hide the transport layer from the
network devices (e.g. from home routers) then they can't enforce the
single path assumption. In fact, once we disassociate location
(addressing) from connection endpoint identification (like described
in TOU) then connections should be able survive even across an address
change and between two completely providers. This is of huge value to
our users and IMO justifies encrypting the transport layer.

Tom