Re: Idea for algo.

From: John Savard (jsavard_at_ecn.aSBLOKb.caNADA.invalid)
Date: 12/31/03 

Date: Wed, 31 Dec 2003 04:43:41 GMT

On Tue, 30 Dec 2003 22:48:02 -0500, "Douglas A. Gwyn"
<DAGwyn@null.net> wrote, in part:
>Peter wrote:

>> Most algos that I've seen involve the use of XORing, addition,
>> subtraction, rotation & bitshifting in any combination and to various
>> levelsof complexity.

>Really? All the ones I've seen just use NAND (apart
>from a few that also need a true random bit source).
>:-)

Note the smiley. ALL logical operations can be built up from NAND
gates.

Even so, of course, using lots of NANDs in an inefficient way might
well introduce desirable properties, like nonlinearity, that using
fewer NANDs more directly does not.

This is why DES has S-boxes, for example.

A block cipher where part of a round involves using the f-function of
one-half of a block to generate an S-box to apply to the nibbles of
the other half would be nonlinear in a profound way that is missing
from typical block cipher architectures. This would not be a bad
thing, and would make analysis more difficult, in a way that just
piling on more rounds might fail to.

John Savard
http://home.ecn.ab.ca/~jsavard/index.html

Relevant Pages

  • Re: Simple block cypher for 8-bit microcontrollers
    ... Well, block cipher design and validation is very expensive, if you ... one round of the algorithm. ... This would be a kind of hybrid differential/slide attack. ...
    (sci.crypt)
  • Re: [PATCH] OpenBSD Networking-related randomization port
    ... >> world needs a small block cipher for some of these applications.) ... TEA is intended for the latter niche. ... if I take a k-bit counter and encrypt it with a k-bit ... invertible) round functions f. ...
    (Linux-Kernel)
  • Re: Rijndael/Blowfish Cipher Question (correction)
    ... cryptographically strong and nothing else is the block cipher in our ... > that in the present context two rounds instead of three ... P1 (which is a requirement for the block cipher that is fullfilled by ... After the second round every bit of P1' depends on every bit of the ...
    (sci.crypt)