c-c-c-combo breaker

It's not often that I brag about my failures in advance, but I thought I would run a little something up the pole here and see if anyone salutes. It looks like a good part of this weekend will be spent trying to decipher the John Brown family's encryption. Probably unsuccessfully.

When John, Jr*., wrote to his wife Wealthy, if he had something he did not believe it safe to write out for whatever reason, he would encode it like this:

12 03
8 02
24 18
25 01
25 22

I have already identified a number of patterns, such as the fact that the pairing 00 08 appears often (and always apart from other groups), that the first column runs from 00 to 25, but the second 01 to 26**, that the periods at least appear to be used as periods, that spaces only appear in column 1, and that the reading is columnar rather than linear. Tonight after everyone is in bed I shall grab a glass of wine and a pencil and really go after it.

However, if any of you pattern recognition geeks would care to take a look and throw some ideas my way, I would be appreciative. I would not really care about the code, except that the correspondence in which it appears is written during a period (Jan., 1862) in which it is vitally important for me to know what Brown was truly thinking about a number of things.

* and perhaps others, I just have not seen it elsewhere.

** Which spread leads me to believe we are dealing with letter-by-letter encryption(perhaps on a matrix), which is breakable, rather than a page/book scheme, which is hopeless.

UPDATE: SUCCESS! And failure. I worked on it until I fell asleep last night, then again this morning.

It's not a matrix, but a straight-line letter substitution.

I started off by graphing the occurrence of each number by column, but that really didn't seem to get me anywhere and as I discovered a matrix would not provide enough letters needed per word, I quickly discarded the idea that the column itself was important*.

I also noticed that where there were spaces, they always occurred in the left column and that the periods were consistently closer to the lower group than the upper. That lent credence to the idea that the periods were punctuation and one spelled the words bottom up.

Then I made the provisional assumption that they were spelled like this:

FE
DC
BA

Which would give words with odd numbers of letters a blank in the upper left position.

Then I concluded that the reason 00 08 always appeared in isolation was that it represented "I" - the zeroes being inserted to mask the common occurrence of 08 by itself.

Then I graphed the occurrences of each number on the middle page and compared the result to a chart from wikipedia of the overall occurrences of letters as they appear in everyday writing.

Then I located a number of recurring number groups with three and four letters - these were obviously important and/or common words that could be used for testing or process-of-elimination.

Then I isolated all the 2-letter groups and compared them to common 2-letter words (e.g. am, an, as, at) to isolate potential vowels.

Then marking a few that looked like vowels, and isolating the most common consonants, I wrote out the alphabet and put prospective numbers under them.

It took about 5 numbers before a pattern emerged and it proved to be correct: start with A as 24 and reduce by 2 (B=22, C=20) until you get to 2, then start the next letter (N) at 1 and go up by 2. Y=25, and it doesn't matter what Z is, as there were none.

My common words quickly proved to be "love" "that" and "you," and the only thing that threw me off was the word "Weppy" in quotation marks, which turned out to be Brown's pet name for his wife. I made the mistake of trying that first instead of last.

Decrypting the whole thing from there was simple if tedious.

So that's the success**. The failure is that it turned out to be a mush letter which did not give me what I was hoping to find. In fact, I felt a little guilty reading the guy's mail.

* I also realized that since this was going to his wife and not a Confederate agent, the scheme did not have to be overly complicated.

** ignoring numerous rabbit trails that I thought might be shortcuts, like Caesar offsets and finding 8-letter words like "Jennison" and "Missouri," which would obviously give me lots of letters quickly.