Re: Random Ciphers

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

Date: Sun, 29 Feb 2004 23:04:54 +0100

Protego wrote:

[snip]
> http://www.protego.se/encrypt.htm

For long time, I have repeatedly argued for the advantages
of crypto designs that are 'variable' in the sense of
either having parameters that are key-dependent (but then
fixed, e.g. key-dependent S-boxes) or having operations
that are dynamically variable in nature (in certain limited,
i.e. not weird, sense), e.g. employing the output of a PRNG
or values of processing of the preceding blocks (plaintext,
ciphertest, and/or values internal to the block processing)
to influence the processing of current block. That
influencing could be through e.g. xor-ing the PRNG output
with the values in diverse rounds of processing of the
block, selection of S-boxes from a (eventually larger) set
of S-boxes, dynamically generating (or modification) of
S-boxes, permutation of bytes (or hexs/bits) of intermediate
values of processing of the block, feedback to the PRNG
with values from processing of the block (to influence
future output of the PRNG), permutation of round keys (with
respect to the rounds), and other conceptually akin measures.
In a recent thered, I also suggested employing a sort of
'internal' chaining, i.e. xor-ing the intermediate values
(i.e. of the diverse rounds) of processing of the previous
block (or the previous blocks) to the values of processing
of diverse rounds of the current block.

I could only see in your article that you also advocate
the basic idea of having 'variability' (dynamics) but
without mentioning any concrete measures of doing that
and I don't see why you consider the idea presented in
it to be new/novel at all. Note that I am not the single
one propagating 'variability'. A few others in the group
have also from time to time expressed positive opinions
of having 'variability' in crypto designs before.

M. K. Shen
---------------------------------------
http://home.t-online.de/home/mok-kong.shen