RE: VPN Glossary On Line!

[Rick Smith <rick_smith () securecomputing com>]

At 09:14 PM 01/10/2000 -0600, Tina Bird wrote:
> The yearly energy output of the sun is a unit in common use
> in astrophysics (never mind the solar-system-centric point of view).
> ....  But people do frequently
> refer to the length of time to search a particular key space in
> terms of multiples of the age of the universe, so it's not much
> of a stretch.

What we need is a generic way of measuring the resources required for
cracking keys. 

Time, as in "age of the universe," is one way to measure resource
requirements, but most attacks can trade that off against improved
computational speed or parallelism. "Yearly output of the sun" ties the
resource estimate to energy, which is the ultimate measure of resource
requirements.

The "annual solar output" argument says there's a "minimum amount of
energy" required for any computation. That provides a nicer basis for
estimating strength, because it encompasses the trade off between time,
parallelism, and improving computational speeds.

Schneier provides an easy-to-find version of this argument in "Applied
Cryptography" under "brute force attacks" and "thermodynamic limitations."
Essentially, the second law of thermodynamics establishes the minimum
amount of energy necessary to change the "state" of a particle. If you
collect all the energy from the sun over a year's time, that will cycle the
state 2^197 times.

So, arguably, it will take *at least* that much energy to test every
possible value of a 197-bit key, regardless of how many computers you use
or how fast they are. The limitation is in our ability to collect immense
amounts of energy and channel them into a computation, regardless of how we
structure the computation in space or time.

This all works as long as the notion of entropy in physics/thermodynamics
is really and truly the same as entropy in information. Intuitively it
makes some sense, so I'm more inclined to see it as a "speed of light" sort
of limit. This has some interesting implications regarding Moore's Law, too.


Rick.
smith () securecomputing com
"Internet Cryptography" at http://www.visi.com/crypto/