
From: Scottauge@aol.com Date: Tue, 20 Aug 1996 18:36:45 -0400 Subject: cryptoanalysis 002
For example, if the crypto-alphabet for e is 23, 45, 190, 200, etc, one can remove some of the frequency for a letter. This definately makes it harder to attack with the frequency analysis method because the "resolution" of the distribution for the letter is lessened to a near randomness. (So it looks, there are still clues, eh?)
I think (from memory) this is called "homoalphabetic". Encyclopaedia Brittanica (Cryptology - article by Gus Simmons) says that it is still vulnerable to frequencies of digraphs, trigraphs etc. But even Gauss was keen on it once. I guess it might have value as a part of another system, making a known plaintext into one of many. Whether it's worth the increased cyphertext size in a system you'd hope to be immune to known plaintext attacks anyway is another question. PA