Re: CAST self-decrypting archive; RC6

From: Mok-Kong Shen (mok-kong.shen_at_t-online.de)
Date: 02/29/04 

Date: Sun, 29 Feb 2004 12:59:31 +0100

Gregory G Rose wrote:

> RPK <webmaster@hypercypher.zzn.com> wrote:
[snip]
>>Some of the algorithms like RC6 which support 2040-bit keys are often
>>not seen, whereas RC4 is still popular.
>
>
> RC6 is patented. RC4 is not secure by today's
> standards.

The paper

   http://research.microsoft.com/users/mironov/papers/rc4full.pdf

says that discarding some multiple of 256 bytes of the
initial output would be a good remedy, if I understand it
correctly. (BTW, repeating the keyscheduling a few times
apparently would have the same effect of rendering the
initial state of the S-box sufficiently random.) I believe
that employing a third counter (which affects both the
permutation in the S-box and the output value via a third
S-box value taking part in the modular addition) would very
likely give essential amelioration, though with a trade-off
of speed. (See the recent thread 'A variation of RC4').

M. K. Shen