Threats in real life - what are we worried about?

[Matt Blaze <mab@research.att.com>]

Tim May writes:
> TEMPEST has very little to do with Cypherpunks goals, actually. First,
> buying such a gadget, tweaking it, exploring capabilities, etc., would
> lead to what? The ability to park a van in front of someone's house
> and--maybe--monitor their screens? We already know this is possible.
> (You all knew that, didn't you?)

I disagree.  "TEMPEST" risks and countermeasures are but one entry on
a long list of subjects in which our ignorance (and that of the
civilian security community in general) may well come back to bite us.
Granted, this is the "cypherpunks" list, not the "securitypunks" list,
but it behooves anybody interested in developing strong mechanisms to
accomplish some security objective to be at least acquainted with how
those mechanisms fit in the larger picture.

One of the most dangerous aspects of cryptology (more dangerous,
perhaps, than fact that there are almost no solid theorems that tell us
how secure practical ciphers really are) is that you can measure it.
It's all too tempting to misuse an estimate for the cryptographic work
factor for some cipher as if it were some kind of overall security
metric for the systems in which it is deployed.  In real life, there
are lots of ways to violate system security, including cryptanalysis,
protocol attacks, Trojan horses, viruses, electromagnetic monitoring,
physical compromise, rubber hose cryptanalysis, OS bug exploitation,
application bug exploitation, hardware bug exploitation, user error
exploitation, physical monitoring, social engineering, court orders,
dumpster diving, and so on and so on.  Most of us on the list like to
think about cryptography and cryptographic protocols, and that's fine,
but it isn't the same as thinking about building secure systems that
are strong enough to withstand attackers who aren't willing to
restrict themselves to a strictly cryptographic threat model.

Unfortunately, the world outside the cypherpunks list isn't much
better off than we are in understanding these "informal", but all too
real, threats.  Thinking about some of them would, I think, go a long
way toward contributing to "Cypherpunk goals" as I understand them.
(Practical TEMPEST shielding is one such problem.  Another good one is
the almost completely ignored problem of storing secret keys on
networked computers.  Still another is the problem of using security
software remotely with limited local computation.  There are lots
more.)

That said, no one can force these discussions to happen, and no one,
much less me, has a right to complain that everyone else is talking
about the "wrong stuff".  So let me raise a question:
	
	Given existing crypto tools (PGP, etc), what are the top ten
	practical attacks against the privacy of stored data and
	electronic mail?  Who are the bad guys? What tools do we need
	to limit these threats?

I'll post my own thoughts later.

-matt