20120619, 09:05  #1 
Apr 2010
Over the rainbow
5×11×43 Posts 
Fujitsu cracks 278digit crypto
http://www.theregister.co.uk/2012/06..._world_record/
Code:
We were able to overcome this problem by making good use of various new technologies, that is, a technique optimising parameter setting that uses computer algebra, a two dimensional search algorithm extended from the linear search, and by using our efficient programing techniques to calculate a solution of an equation from a huge number of data, as well as the parallel programming technology that maximises computer power. http://www.fujitsu.com/global/news/p...12061801.html Last fiddled with by firejuggler on 20120619 at 09:06 
20120619, 09:45  #2 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2
10001101110110_{2} Posts 
"21 personal computers (252 cores) in 148.2 days?" (The 0.2 days precision looks quite out of place.) Sheesh...

20120619, 09:56  #3 
Apr 2010
Over the rainbow
5×11×43 Posts 
Now it's time to generalise 2048bit crypto...
even if a wrench might work far faster... http://xkcd.com/538/ Last fiddled with by firejuggler on 20120619 at 09:57 
20120619, 09:57  #4 
Nov 2010
32_{16} Posts 
It probably has nothing to do with factoring. More like with DL over some exotic group or EC.

20120619, 12:01  #5 
(loop (#_fork))
Feb 2006
Cambridge, England
18D1_{16} Posts 
http://www.nict.go.jp/en/press/2012/...20120618en.pdf is a PDF describing what they've done: there's an indexcalculuslike attack which gets you from an EC problem over GF(3^97) to a discretelog one over GF(3^582), and they've then done that one using the function field sieve.

20120619, 12:10  #6  
Oct 2007
3×5×7 Posts 
Quote:


20120620, 06:31  #7 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2
2×3×17×89 Posts 
So it has asymptotically equivalent difficulty to SNFS? (Not to GNFS?)
Why did they compare it to the 676bit GNFS factorization? (I am only echoing Tom. I am not that smart. Also.) Last fiddled with by Batalov on 20120620 at 06:35 
20120620, 10:31  #8 
(loop (#_fork))
Feb 2006
Cambridge, England
6,353 Posts 
I think they're comparing it with the functionfieldsieve in GF(3^426) (a 676bit number) which an NICT group did and described in http://eprint.iacr.org/2010/090.pdf. The press release doesn't mention factorisation.
The unisaarland link is for algorithms in large prime characteristic (though both generalprime and specialprime), which are rather different from the function field sieve that is used in small characteristic. Last fiddled with by fivemack on 20120620 at 10:36 
20120620, 20:03  #9 
"Nancy"
Aug 2002
Alexandria
4643_{8} Posts 
Moved from Factoring to Science since it's not a factorization problem they solved.

Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Crypto News  Nick  Tales From the Crypt(o)  51  20200211 23:37 
ElGamal crypto without prime  ElChapo  Math  9  20170610 03:26 
25GPU cluster cracks every standard Windows password in <6 hours  swl551  GPU Computing  4  20121222 01:32 
Crypto 2007  R.D. Silverman  Lounge  2  20070808 20:24 
crypto game  MrHappy  Lounge  0  20050119 16:27 