On 20/09/15 14:53, Georgi Guninski wrote:
Found this from a DJB paper:
http://www.scs.carleton.ca/~paulv/papers/JoC97.pdf
Parallel Collision Search with Cryptanalytic Applications
Paul C. van Oorschot and Michael J. Wiener
CHECK THE DATE:
1996 September 23
Both authors are well-known.
Google says the paper was published in the Journal of Cryptology in 1999.
p.1
The practical significance of the technique is illustrated by giving the
design for three $10 million custom machines which could be built with
current technology: one finds elliptic curve logarithms in GF(2^155)
thereby defeating a proposed elliptic curve cryptosystem in expected
time 32 days, the second finds MD5 collisions in expected time 21
days...
The present day open ECC dlog record stands at about 114 bits, iirc:
that method used ~2014 custom hardware, but not $10 million worth.
I'd guess Oorschot and Wiener got something in the numbers wrong. It
happens.
However the parallel collision search technique they describe is very
real, and has been used to effect. At a guess, the ECC dlog record above
probably used it, as will most modern collision search algorithms.
As DJB quoted them, I'd guess that they invented the technique (though I
knew of the technique, I thought Knuth described/invented it).
It's one of those things which are obvious in hindsight; but which can
be dev'lishly hard to come up with in the first place.
-- Peter Fairbrother
---
I know the dollar is not what it used to be, but same applies to
hardware IMHO
Metadata of the PDF is in the future, suggests windows.
This is paywalled:
http://link.springer.com/article/10.1007%2FPL00003816
Journal of Cryptology
January 1999, Volume 12, Issue 1, pp 1-28
