[Cryptography] Unicity distance of Playfair

[Ray Dillinger]

On 03/26/2016 07:10 AM, mok-kong shen wrote:

> As is apparent from my OP, my knowledge of unicity-distance is very
> poor. Thus I continue to wonder whether one could in computing the
> unicity-distance of a scheme the fact that in it diagrams are being
> processed in each step instead of single characters. For, as a
> parallel, in cryptanalysis one could study not only single letter
> frequency distribution of the alphabetical characters but also
> additionally the frequency distributions of digrams, which provides
> correspondingly additional informations.

The redundancy is an estimate of how much information is available
from all those sources taken together.  Distribution, pairs, triples,
word frequencies, the whole business.  English conveys about 1.5 bits
per letter - the remaining representation bits, ie, the redundancy,
are 'redundant' because that information is both explicit in those
bits and implicit in that you could predict them on the basis of
all that other information.

> .... Take the
> case of the classical poly-alphabetical substitution with a 26*26 table.
> That table could have columns that are sufficiently random on the
> one side (normally termed "with mixed alphabets") or have columns that
> are simply shifted versions of the the alphabet in its natural order
> (normally termed a Vigenere table) on the other side. Now the cipher in
> question is evidently stronger in the first case than in the second
> (special) case. However, the unicity-distance would be the same for
> both cases.  Am I right in this?

I don't think so - a mixed-alphabet Alberti cipher seems a *lot*
harder when solving it by hand.  I believe the mixed-alphabet
multiplies the number of different keys effectively. I've never come
up with a solution to one that used a mixed-alphabet different
from the one the encryptor used, and if it were redundant I think
I would have.

But it often happens with classic hand ciphers that a mixed-alphabet
substituted for a plain one doesn't actually increase the number
of possible keys.  For example when you compose two monoalphabetic
ciphers you don't get a cipher with 26! times as many different
keys (and longer unicity distance), you just get a cipher with
26! different ways to write down each key (and the same unicity
distance).

				Bear



-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: OpenPGP digital signature
URL: <http://www.metzdowd.com/pipermail/cryptography/attachments/20160326/d14fa7c8/attachment.sig>