Re: A revised "FFT" hash function.

From: Tom St Denis (tomstdenis_at_iahu.ca)
Date: 10/03/03 

Date: Fri, 03 Oct 2003 12:27:56 GMT

I've revised the "design" yet again [disclaimer still applies] to improve
the sbox.

Changes

1. I use a different mixing network for the sbox. It has a comparable
delay but gives much better results [3 round feistel was almost always with
a NonLinearOrder [NLO] of 5 not 7]

2. The delta [DPmax] is now 8/256 not 16/256.

3. The lambda [LPmax] is not 30/256 not 32/256 [not a huge change :-(]

4. The NLO is now 7 not 5.

I've also lowered the round count to 10 because I strongly disbelieve that a
DC/LC attack could make it through 10 rounds [the best possible
[theoretical] 8 round approx would have in excess of 2^240 work]

Mads has expressed interest in "papering" up the design [I wrote it for him
originally anyways]. So hopefully he/we'll have a formal presentation up.

http://iahu.ca:8080/src/ref.c

Tom

Relevant Pages

  • Re: polk 6x9
    ... He started out trying to prove it by claiming "if they were any good you'd see home speakers use them". ... it is no harder to manufacture an oval cone than a round one. ... And design software these days - like Finite ELement Anaylsis - makes it no harder to design, ...
    (rec.audio.car)
  • Ibraheem al-Ehabi might scatter Neal
    ... My impossible breeze won't design before I ... manipulate it. ... round the constant avenue were finishing of the common foothill. ... While cables truly carve architects, ...
    (sci.crypt)
  • Re: MC2
    ... AES wasn't the first cipher to use inversion [also known as ... You want a sbox with a good hardware footprint, ... make the design harder to analyze. ... makes the key schedule resistant to related key attacks]. ...
    (sci.crypt)
  • Re: A basic question about hashing
    ... Without seeing a concrete design it's a bit hard to come up with an ... you have to use some feedback to process multiple blocks. ... IIRC Merkle wrote a paper in the eurocrypt series on making hashes ... MD5 et al were broken because the round transform ...
    (sci.crypt)
  • Re: a new very fast hash algorithm (160 bits), with a technique of "overlapping sums"
    ... >> this design? ... > used as hash function, as a block cipher or as a stream cipher. ... Look at the mechanical operations in the round() ... > integer of the block before the rotation. ...
    (sci.crypt)