Thanks for detailed and useful comments!

On comment 2: I ran your updated ASCII files of Exhibits 1 and 4 through my code and they both decipher without any transcription errors now. We can only hope that the typesetters did not make too many errors in Exhibits 2 and 3.

On comment 3: My explanation of how to get the so-called ‘cipher-text’ key was incorrect. I had originally done as you suggested and simply derived the cipher-text key from the plain-text key. I updated the article to make this clear.

On comment 4: Agreed, the use of agnostic terms to refer to disks/rings are better: article updated.

Thanks again,

Carl

FYI I added a link to your article on Wikipedia’s Chaocipher page. Feel free to edit it in any way you see fit.

Moshe

You’ve done a fantastic job, complete with clear diagrams and description — true “eye and mind candy”! I am planning on posting links to it from The Chaocipher Clearing House.

Some minor comments:

(*) For the record, Byrne’s original mechanical model used two non-concentric disks, with the disks being engaged along the edges, and turning one turns the other (when engaged) as seen on the National Cryptologic Museum web site. Nonetheless I, like you, found it easier to describe a slightly different but equivalent model.

(*) You warn would-be solvers of transcription errors in Exhibit 1′s ciphertext text file on TCCH. Just the other day I realized that, although I posted a list of the nine typos originally found in the TCCH ciphertext text file back in June 2009 (due to OCR errors), I had forgotten to upload the corrected file to the site (mea culpa!). As of yesterday, the plaintext and ciphertext files of all the exhibits are cleaned and authorized for use. You may want to remove the warning section, as it is no longer relevant (please double-check me!).

(*) In your pages you bold numerous phrases. The letter coloring (a light blue) is identical to that of links (without the dotted underline denoting a link). I found that a bit confusing. Would you be able to bold in black to avoid the visual confusion?

(*) If I understood correctly, you find the second candidate keyphrase (“THIKKTBNDB”) by running your keyphrase enumerating program again, this time using the left-hand alphabet for locating the ‘plaintext’ letter.

An alternative way of arriving at the second candidate keyphrase “THIKKTBDNB” is by just running the first one, “TILNOYHIVK” through Chaocipher in the standard way (i.e., using the right-hand alphabet to locate the ‘plaintext’ letter), with the final ‘ciphertext’ being “THIKKTBDNB”.

(*) In the “Results” section you refer to the plaintext and ciphertext disks (“It is also possible that the key was not meant to be applied to the plain-text ring (i.e. enciphered), but to the cipher-text ring (deciphered)”).

A reader may be confused when directed to find a plaintext letter on a ciphertext disk. In my original paper I refer to a “right-hand alphabet” and “left-hand alphabet”, thereby decoupling the disk from being pt or ct. You might want to consider referring to the disks as “inner” or “outer” or some other pt/ct-agnostic description.

All-in-all, you’ve done a most excellent job. I look forward with great interest to your upcoming articles!

Best regards and kudos,

Moshe

P.S. I have a paper (to be uploaded any day now) describing my deciphering of Exhibit 1, which also includes some historical background, etc. But your article is true “eye-candy” — wonderful to read !

My write-up of breaking Exhibit 1 is finally done (with lots of pretty pictures!). The next installment will be on breaking Exhibit 4.

http://www.inference.phy.cam.ac.uk/cs482/projects/chaocipher/

Any comments or corrections welcome.

Carl

Many thanks for that original direction! Although Byrne probably did not contact CSEC directly, they may have tried to solve Chapter 21 on their own. Certainly worth a try!

I was not able to find any reference to “freedom of information” etc. regarding CSEC. Nonetheless, I will schedule this as a task for the near future.

Thanks!

Moshe

Can you already place this and the “Chaocipher got Slashdotted” entries under the “Chaocipher” reference page? They’re about to scroll off the main page … !

My algorithm iterates through the pt/ct, starting with offset 0, and computes the # pairs needed from that point forward. It keeps track of the best match, i.e., the one requiring the fewest holes. Could you please tell me what your pt/ct was that gave only four holes? I’d like to see why my program (a relatively simple one) didn’t catch it.

Thanks,

Moshe

Indeed, I followed very nearly exactly the same method as in your post: searching for the sequence of characters with the fewest holes and searching through the possible ways of plugging the holes. I see in your post that you found the smallest number of holes to be 5. With a very slight modification in the method, it is possible to do it with only 4 holes. This is not so important for Exhibit 1 (a very long text) but could be important for shorter texts where we have a smaller chance of finding good sequences of characters.

In my version, I search not only forward in the text for characters that have already been seen, but also backwards. This provides a better chance of filling in more of what we already know about the alphabet. To search backwards in the text it is necessary to unpermute the alphabets, i.e. make the opposite permutation of what is required for searching forwards in the text. Other than that the search remains the same.

My only remaining secret is how to find the keyword to the cipher from the initial alphabets. I’ll describe that soon when I write up the rest of the method–hopefully between tomorrow and the weekend.

Then there is also the problem with Exhibits 2 and 3. As you wrote, they do not follow the same statistics as standard Chaocipher texts. I have tried out some pet theories on how they might have been modified, but no luck so far.