[3459] in cryptography@c2.net mail archive
RE: "It's a Hardware Problem..." (fwd)
daemon@ATHENA.MIT.EDU (decius@ninja.techwood.org)
Wed Oct 14 02:00:56 1998
From: decius@ninja.techwood.org
Date: Wed, 14 Oct 1998 00:22:14 -0500 (CDT)
To: cryptography@c2.net
> "World's Smallest Combination Lock" Promises
> To Foil The Best Computer Hacker, Say Sandia
> Developers
I like this general idea, but is it really necessary to use nano-machines.
I'm pretty sure I could implement such a device purely in logic. Here is a
simple design that came from a few seconds thinking about it. Something
more robust could be implemented with a little work.
Use a bit comparator. The "key" is loaded into the comparator from
eeproms that can only be programmed from the secure side. The insecure
side sends an enable bit after the key has been setup on the comparator.
If the keys from both sides are the same, the comparator sends a high
voltage to a multiplexor, which connects the lead between both sides of
the system. The multiplexor input can be held high by a flipflop for a
given amount of clock cycles or until the enable bit goes low again or
whatever you want. If the enable bit for the comparator goes high and the
bit comparator does not give a "true" output on the next clock cycle then
the "tripped" flipflop is turned on, which sends a zero thats anded with
the enable bit as it goes from the insecure side to the comparator
(nominally this would be a one). The "tripped flipflop" can only be reset
from the secure side. This could be set up with a "secure side
comparator" which will only allow the "tripped" flipflop to be reset if
the user on the secure side has the right key. Keep in mind that this
whole thing would be inside an ASIC.
I could cook up VHDL for this and send to an ASIC fab in Tiawan and be
ready to ship the chips before Sandia goes to market. Its got to be
cheaper than fabbing nanomotors. I'm probably not violating their patent
as my implementation is totally different. Anyone want to invest a few
thousand on a chip run? :)
One note, someone is going to argue that the sandia chip is secure from
electromagnetic attack because the code is stored mechanically. However,
one would think that changing a value in an eeprom with an electromagnetic
attack would be a hell of a lot harder than causing an electromagnetic
motor to turn (or any number of other possible attack scenarios). The
kinds of situations where that will be the deciding issue are severly
limited.
Am I missing something here? It is rather late.
Tom Cross
