| Viewing Single Post From: Brute Forcing The Adfgvx Cipher | |
|---|---|
| jdege | Apr 28 2008, 04:27 PM |
Elite member
  ![]](http://209.85.122.85/static/1/pip_r.png)
|
IFF you have a good test for whether the result of a given trial transposition looks like the output of the substitution phase. The first step is a fractionated substitution, creating two letters for one. But it's a simple substitution, and a digraph frequency count would show a normal language distribution. For that matter, converting it back to single characters using any arbitrary Polybius square would yield a result that was only a simple substitution away from the plaintext. So there's no question that there are statistical measures that distinguish the output of the substitution phase. The question is whether the output of the trial transpositions are statistically distinct. It's clear to me that most of them will be. The majority of incorrect transpositions will yield something like a random shuffle, that would result in a digraph frequency distribution very different from plain text. But I'd not feel in comfortable in stating that all of the incorrect transpositions will be so. It seems perfectly possible to me that some number of incorrect transpositions will look correct. Whether this is true, or how many of them there are, I'd not even pretend to say, but it seems to me that the number of transpositions that could be correct is the major driver as to whether this is a viable approach. Of course, given multiple messages encoded with the same key, things are very different. While it might be possible that a number of incorrect transpositions yield normal-looking frequency distributions, the odds that the same incorrect transposition would yield normal distributions are very low. Still, you've been playing around with this. How many false positives have you been seeing? |
| When cryptography is outlawed, bayl bhgynjf jvyy unir cevinpl. | |
 |
|
| Brute Forcing The Adfgvx Cipher · General | |
4:08 PM Nov 25
