Date: Sun, 29 Jun 2008 13:26:22 -0400
From: Glenn Everhart <Everhart@....com>
To: full-disclosure@...ts.grok.org.uk
Subject: Let's design a spy-proof communications
	infrastructure

Colleagues:

It is unworthy that people should be spending energy criticizing
others' qualifications, personal habits, ancestry and destination (as
the wording goes).

I suspect that something much more useful can be possibly facilitated
here (and elsewhere if anyone feels like it).

Let me suggest that it should be possible to construct something like a
cell phone network which will run like a peer to peer network, with
routing determined heuristically and pretty much unpredictably, with
message encryption, and with small enough electronics to package
in something no larger than current cell phones.

The current designs we have are the creatures of the old phone companies
and presume things go through central offices. This has led to intrusions
into user privacy by crooks and governments, and tends to make all manner
of information we might not care to publish become effectively wide open
to anyone who cares to steal it.

However, consider that many internet p2p networks have been worked out
(and are still being) to hide some of this. Consider that the old usenet
protocol has no idea of global source or destination (though its flood
fill algorithm is I suspect way too slow, still, to be used for messaging
or voice traffic). If a network is designed so that every member only
has some idea of its neighbors and which of them might be closer to the
desired endpoint than it is, each node only has or needs a very local
idea of addressing - something that might be relatively useless to
central authorities or to crooks.

The electronics to receive and send messages locally can be made very
small and cheap. There are low power CPUs from places like TI and Atmel
that run on microwatts, and WWV receivers can be had for $1 in chip
form in bulk (per messages I have gotten). We have GPS boxes that
you hold in your hand able to receive satellite transmissions. A few
years back this would have been thought energetically impossible.

If we devised some private communicator, it might expect to function
in a very large net so long as some path existed to other communicators.
While truly global routing might require some relays to bridge areas
with few people, in urban areas and quite a few not-so-urban ones direct
communication should be workable, at low enough power on any single
frequency
(yeah, make it spread spectrum) that formal licensing would not be needed.

It should be noted that the address of any such system need not be
fixed for huge times. To the extent you can get the systems to read,
say, a time synchronization signal, systems might simply pick new addresses
out of a suitably long number space. (If this is truly random, address
collisions might be made so rare they can be ignored.) This would
mean routing would need to be recomputed locally every so often but
would make the notion of global address pretty well meaningless and
unpredictable. (Use a heat source perhaps to generate random bits, so
the randomness is from thermal noise. Nobody will be able to steal
a key and figure the next address, or the last...). If a broadcast were
available so each unit could sense nearby ones (where you make "nearby"
as far away as you can) the constantly changing addresses won't cause
problems discovering what else exists. If you have to scan an area,
such discovery could be unsecurable. While I mention discovering where
one is on a mesh, this might be tried with and without actual geographic
coordinates. Nearness measured by a Hamming distance could be used for
routing also. It might not be as efficient but if it worked it would
mean routing gave eavesdroppers no hint as to physical location of
anyone. If we want to keep private conversations private, this seems
like a good thing.

Authenticating people is I think separable from this; I have some other
schemes to handle that. For a communicator, encryption should basically
make traffic snooping impossible and make routing snooping infeasible
even with adversaries who listen to a lot of traffic. The lessons of
Blackberry should be heeded here: make the encryption all end to end,
not step by step, with no backdoors built in and with open source code
so tampering with these principles can be quickly caught and negated.

Building such gadgets would be paid for by people wanting to use them,
but note that the necessary infrastructure is just the existence of
a large bunch of these things being used, sitting on peoples' belts or
in pockets and passing traffic among one another. You start selling
them in small offices or families, where the necessary groups will tend
to be together a lot. Gradually people will notice that they can
reach others.

How to address some particular person then?

I would suggest that some of the p2p research might be useful here.
Perhaps have the gadget transmit a name or other identifier of the
person there in some form. If for example we allow repositories
of public keys, we might transmit "John Smith has address xxxxxxxxxxxx"
where xxxxxxxxxx is encrypted by his private key. (This is not very good.)
If a few trusted nodes can be made, they can be used in setting up
connections by finding the current addresses.

If I wanted to talk to John Smith and could find a partial address
for him from some repository I might transmit some of my
address encrypted by his public key and my name, and it might
be noticed by John Smith's communicator and full address sent
back. This kind of thing gets somewhat better, since not everything
gets sent at once. It is still not great.

It is probably best for routing to have all units be able to be
opportunistic
routers, so that there would be a large and often - changing set of
routers in

any area having information about some addressees. You will need some way
to convert an identifier "name" to a network address, probably in several
pieces to make it hard to fake.

It is probably worth thinking about a "web of trust" here and having ways
to declare identifiers as trustworthy or others as to be shunned. Phil
Zimmermann wrote much about this in conjunction with pgp, and such
function may be necessary in some places to keep eavesdroppers from
hammering parts of the network to try to analyze it. Here too I would
suggest that distributing the decision function could make it harder
to subvert.

This is all pretty schematic, but if a private communicator like this
were devised, our networked conversations might be able to again be
private as they were years back when you just went out behind the barn
to speak privately, and could be pretty well sure nobody else heard.

I think a world in which it is harder for your every move to be tracked
will be harder for anyone to take over and will tempt people less
who now think they can watch your speech and predict you might do
something they dislike.

However please if anyone wants to discuss this, I would ask that
priority be given to what technically can or cannot work, what
actually might protect or what will fail and perhaps be a boon
to eavesdroppers, how much stego is needed in here, or other such
topics. I suspect there is enough technical savvy around now to build
something along these lines (and having said so in public, I may
be said to have let the only important cat out of the bag). Anyone
want to add or detract?

(The foregoing is not un-holey and certainly not all that would be
needed.)

Can people here propose something that is still better?

Glenn Everhart


_______________________________________________
Full-Disclosure - We believe in it.
Charter: http://lists.grok.org.uk/full-disclosure-charter.html
Hosted and sponsored by Secunia - http://secunia.com/

Powered by blists - more mailing lists