Re: "Strong" crypto and export rule changes.

[Ian Goldberg <iang@cs.berkeley.edu>]

-----BEGIN PGP SIGNED MESSAGE-----

In article <199702030626.WAA14616@toad.com>,
Vin McLellan  <vin@shore.net> wrote:
>	Ian popped the 40-bit RC5 (not RC4) challenge with 259 processors,
>almost all standard Unix college-lab workstations, as I understand it.
>(RC5 has a variable block size and a variable number of rounds; but the
>unknown plaintexts for this contest were enciphered using a declared
>12-round RC5 with a 32-bit word size.) The message Ian revealed was
>something like: "That's why you need a longer key!!!!!"
>
>	 (The network Ian used to link his lab workstations, NOW at
>Berkeley, is definitely not standard, however.  I think there is a
>description of it online; but briefly, NOW seems designed to very
>efficiently handle this sort of intensive distributed processing project.
>More important, perhaps, was the fact that Ian just chewed through the
>possible keys with a pure brute-force attack on the key space.  His attack
>was not really optimized for RC5, or designed to attack any specific
>element in the RC5 crypto architecture.)

Actually, it was 259 _machines_, but 4 of them were 8-processor UltraSPARCs
(by far the coolest machines I had access to), for a total of 287 processors.

The "special" network used by the NOW cluster was irrelevant; I didn't use
it at all.  I used the machines on the NOW and 120 other HP workstations
as regular TCP/IP clients.  I started them by simply logging in to each
machine, one at a time with "rsh" (well, "krsh" or "ssh" where appropriate).

   - Ian

-----BEGIN PGP SIGNATURE-----
Version: 2.6.2

iQCVAwUBMvj4REZRiTErSPb1AQFl5gP+MGlyElNu6X3IHseW6Q0EPicPa4mQs35Z
koUKkAhk0qrT2CpEzw7J6dtjyTLs2BUmScEOtvU8KiBjK8aRZCsE0BHSmONWtX71
dNZu1q/+wm2oSLi1tDq0mT7bpbBR0NbO71tWgza2vTFhtP4vKvzt5SodYSN+JTYL
5DuLvpofRFs=
=ly1x
-----END PGP SIGNATURE-----