Talk:EFF DES cracker

From Wikipedia, the free encyclopedia

Contents 1 Chips 2 Controversy 3 Number of custom chips contradiction 4 When is it done 5 Sun-4/470 6 Footnote 1 under References.

[edit] Chips

What did the custom made chips do? Why didn't they use regular cpus for their tasks? --Abdull 23:29, 3 October 2005 (UTC)

General CPUs are much slower than dedicated custom processors, since they are designed to be flexible and generic. With a general CPU, you must issue several hundred instructions just to check one key before you can start the next one (ie: one key per several hundred clock cycles). With a custom processors, it will be limited to only running that specialized instruction, but it can have an entire processor pipeline set up to checking an entire key. Each stage of the pipeline would connect solely to the next stage of the key check, and several keys may be in progress simultaneously (ie: one key per few clock cycles, with many keys in the pipeline).-- Bovineone 01:30, 4 October 2005 (UTC)

[edit] Controversy

I seem to recall that EFF took some heat at the time from those who felt there were better uses for $250,000 than to demonstrate the vulnerability of an encryption standard already widely considered insecure. I don't want to add language to this effect in the article without a noteworthy citation, which I haven't found yet. Does anyone else recall this controversy? I think there was some crossover with those who felt Barlow took the group the wrong way on Digital Telephony.--Skyraider 00:46, 3 February 2006 (UTC)

I don't think there was that much controversy, just some slight skepticism here and there about whether the effort was worth it. I also don't have the sense that EFF paid the $250K out of membership dues and small donations, that might instead of have been used for something else. Rather, my impression is that one of the EFF's wealthy founders (JG) funded the project pretty much by himself, because he wanted to see it happen. Phr 06:40, 12 April 2006 (UTC)

I read that it was several million keys per second, not billion. Also my friend's book stated $200,000 not $250,000. Someone with more resources should verify these numbers...

It was definitely billions of keys per second. The machine could search the whole 2⁵⁶=7×10¹⁶ keyspace in about a week, so that's around 10¹¹ keys/sec. Phr (talk) 05:31, 2 May 2006 (UTC)

[edit] Number of custom chips contradiction

There is a difference of 10x between the EFF DES cracker and Brute force attack image capations each indicating '1800' and '18,000' chips respectively. When searching 1800 appears to be the correct amount, but it would be good to confirm.--ShaunMacPherson 16:54, 15 July 2006 (UTC)

[edit] When is it done

Q: If a machine tries 2^56 keys to decrypt a message, how does it know when it is done? It would have up to 2^56 output strings that would then need to be processed to see whether they seem like the message or gobbledegook.DouglasHeld 21:15, 22 May 2007 (UTC)

After each attempted decryption, the chip checks whether the decrypted plaintext matches the excepted plaintext. If yes, it has found the key, and the crack is over. So, searching the whole 2⁵⁶ keyspace actually takes 2⁵⁵ decryptions on average. -- intgr ^#%@! 20:57, 23 May 2007 (UTC)

[edit] Sun-4/470

Maybe there should be some reference to the machine being built in a Sun-4/470 chasis (VMEbus-based). A minor point, but perhaps of interest. Thanks. 75.142.145.210 (talk) 20:37, 23 April 2008 (UTC)

[edit] Footnote 1 under References.

The first URL appears to be a broken link. 75.142.145.210 (talk) 20:40, 23 April 2008 (UTC)