1996-07-24 - Re: Brute-forcing DES

Header Data

From: koontz@netapp.com (Dave Koontz)
To: tcmay@got.net
Message Hash: 931e1b175119fe416804334f5bd6449ba6c6cbad6f3b7b38706f519fcbfd5b31
Message ID: <9607232216.AA22925@supernova.netapp.com>
Reply To: N/A
UTC Datetime: 1996-07-24 11:30:43 UTC
Raw Date: Wed, 24 Jul 1996 19:30:43 +0800

Raw message

From: koontz@netapp.com (Dave Koontz)
Date: Wed, 24 Jul 1996 19:30:43 +0800
To: tcmay@got.net
Subject: Re: Brute-forcing DES
Message-ID: <9607232216.AA22925@supernova.netapp.com>
MIME-Version: 1.0
Content-Type: text/plain


>>At 3:06 AM 7/23/96, Steve Reid wrote:
>>Did you consider the possibility of DES chips in your back-of-the-envelope
>>calculations? They are hundreds of times faster than PCs. I don't know
>>where to get them or how much they cost, though. I would expect they
>>wouldn't be too expensive. The cash might be better spent on DES chips
>>than on a prize.
 
>Specialized DES-cracker chips have of course been considered.
 
Actually hardware DES can be thousands of times faster than PCs.  The 
problem with using commercially available DES chips is that you need to
load keys, do encrypts and XOR the output with the ciphertext looking
for all 0's or all 1's, in a serial fashion.  Commercial DES chips 
don't have the facilities for doing comparisons or loading a new key while
encrypting with the previous, not to mention the ability to increment the
key value.
 
>The advantage of the cracks done last year, the French and Australian
>cracks, and the MIT cracks, were that the "entry costs" for joining the
>project were low.
 
>--Tim May
 
The lowest cost entry for hardware crackers would probably be FPGA based
(lower NRE).  I could design one that would do say 2 - 4 million DES ops
per second and cost less than 60 dollars (a PCI interface and cheap 
card).  Anyway, hardware cracking can be done on a smaller scale than Wieners 
30 Million DES ops/second, and it could still prove valuable.  
 
The good news is that software and hardware efforts are no more incompatible
than using different performing machines.  If someone steps forward with
1728 Giga DES ops machine, they can have as much of the key space as they
can handle.  I could probably manage a 100 M DES ops hardware machine before
my wife wondered what I was spending the money for the new driveway on (and
I would use reprogrammable FPGAs).  I would also be inclined to run software
on a couple of workstations at home, and an incidental PC or Alpha at work.





Thread