Linux On Steroids: DIY supercomputer software from Sandia

Trei, Peter ptrei at rsasecurity.com
Thu Aug 9 07:47:44 PDT 2001


> Tim May[SMTP:tcmay at got.net] wrote:
> 
> 
> On Wednesday, August 8, 2001, at 02:28 PM, Faustine wrote:
> 
> > Lots of interesting possibilities for cryptographic applications, I'm
> > sure...
> > http://www.cs.sandia.gov/
> 
> 
> Except when was the last time you heard of a Cypherpunks-interesting 
> cipher being broken with _any_ amount of computer crunching?
> 
> (The "challenges" broken by a couple of our own list members over the 
> past several years were all weak ciphers by modern standards, or had key 
> lengths way below even the recommended lengths of the day. [...]
> There are indeed some cryptographic uses for big computers, but not much 
> of real interest here. Some voice- and traffic-analysis stuff, but not 
> cracking modern ciphers.
> --Tim May
> 
I'll second this - the early 40 bit SSL cracks, and the RSA Symmetric
challenges,
operated to force acknowledgement that 40 and 56 bit keylengths were simply
inadequate. I was particularly steamed by some Clinton appointee claiming 
with a straight face that a DES cracker could not be built because "it would

melt down". The success of the challenges created 'facts on the ground'
which could not be hand-waved away by government officials, and meant
that customers started to insist on something better. 

BTW, the distributed.net folks are still cranking at 64 bit RC5. They
have now searched 54% of the keyspace in 1,386 days. 

One thing I'd like to see done with these huge distributed computers
is to factor moduli, such as those in the RSA Factoring Challenges
(the prizes are substantial), A major bottleneck at this point is the
matrix reduction step. If a method could be found to run this in parallel,
then distributed factoring would be greatly eased.

Peter Trei





More information about the cypherpunks-legacy mailing list