Some months after this article was written the algorithm was revealed.
June 1, 2010: The National Cryptologic Museum Foundation (NCMF) acquires
the Chaocipher machine and papers, and on July 2, 2010: Chaocipher is revealed.
See "CHAOCIPHER REVEALED: THE ALGORITHM" by Moshe Rubin.
J.F. Byrne, (John Francis Byrne 1880-1960), was most famous for his
unrelenting declaration of a cipher that was not able to be deciphered, and
for the contest that he promoted in "Chapter 21, Chaocipher," of his book,
"Silent Years: An Autobiography with Memoirs of James Joyce and Our Ireland
(1953)," regarding the Chaocipher machine that he had proposed.
I am one of those people who took a stab at breaking the cipher of Mr. Byrne, and I tried to do this in accordance with what I deemed the best way, which is to say that I intended to try to construct the original cigar box model he brought to the patent attorney, rather than to use sophisticated mathematical investigations of the cipher itself.
I say this, because Mr. Byrne himself did not seem capable of producing the machine he intended, but he was capable of producing the cigar box model. The cigar box model supposedly worked in the office of the patent attorney. My position was that, if the cigar box model could not be re-created, then perhaps the cigar box model that Mr. Byrne used was quite simply re-set each time he made it work for the patent attorney, thereby nullifying any boasts of a code that could not be deciphered.
In other words, if the only thing that mattered was re-setting the start position, the cipher was worth much less than he anticipated. It would amount to "keeping a secret," and any messenger who was captured by the enemy could be made to give up that secret.
I have some authority to say this, because I myself have a circular code in my own invention. My JVP-Encrypt has several methods of encryption, and one of those is the CodeCircle method, which uses letters arranged in two concentric circles. I myself admit that the CodeCircle method is the weakest of the 7 methods I put into JVP-Encrypt. The other 6 methods are much stronger. It would be better if you would code the results of the CodeCircle a second time with another method from JVP-Encrypt on top, to make it truly undecipherable.
First of all, it must be clarified that Mr. Byrne did not have a genuine machine to start out on his journey. He made a contraption out of a cigar box to demonstrate a principle and then sought to have blueprints made of his conception. The quotes from his book spell this out quite clearly.
Second, by all accounts he used a set of wheels or disks in his model and a jumbled alphabet.
Nevertheless, a model is mechanical and this limit must be discussed, since a cigar box and a set of wheels can do only so much when put together in such a state. More complicated contraptions may be thought up, but it is highly unlikely that a great deal of sophistication entered into the plans for the original model.
In other words, if we cannot build a cigar box model ourselves, how could Mr. Byrne build it?
Mr. Byrne's principle, it seems, was that no matter how many times you entered the same plain text, a different message of cipher text would be the result. He wanted random output for every entry, like pulling a ball out of a box, and then returning the ball to the box so it could be chosen randomly again.
In reality, nothing truly random could ever be deciphered, since there would be no way of knowing the precise arrangement of letters in the message. To be able to decipher anything, there must be some underlying motive or method. Therefore, the ultimate desire of total randomness cannot be achieved.
When Mr. Byrne demonstrated his principle in the cigar box to the patent attorney Marcellus Bailey, it was referred to as a toy, and he was advised to not try to patent anything until he had produced at least the blueprints he wanted to produce with an expert draftsman.
As of August 26, 1922, Mr. Byrne was still referring to his cigar box model, although in pages of his book it is referred to as a machine, sometimes with quotation marks and sometimes without. August 26, 1922, is when he was contacted by Major Frank Moorman, who also returned his contraption in a box not well-packed, and thus destroyed.
And, on the same page of the book, Mr. Byrne reflects on the "mistake" he made when thinking he could convince someone of his invention's worth by simply stating the principle. He constructed his cigar box model in 1918, and this was 4 years later, after receiving the message from Major Frank Moorman.
At this point, Mr. Byrne embarks on building a new model, but one more suitable than the original cigar box model he had in 1918. Keep in mind here that he did have the blueprints that he was intent on gaining after visiting with the patent attorney Marcellus Bailey, but the "machine" had been too expensive to build according to the blueprints he had then. He explains this in his book.
When Mr. Byrne receives a reply to his package and letter, from November 18, 1937, it appears that he had sent the document he had produced only as a cipher text, because the reply states that they do not have the personnel to decipher what he sent. With this reply in hand, Mr. Byrne sends another copy, but this time including the plain text from which the cipher springs.
He explains that the first 100 lines all have the same message encrypted, apparently to prove that the same message encrypted 100 consecutive times will be different each time.
There is a problem with the story itself when discussing Mr. Byrne.
First of all, if he regretted getting a patent when he had the cigar box model, he could have filed for one at a later date. He never did. A patent can be gained if you can demonstrate on paper, or with a model, how it works. Mr. Byrne never attempted to patent his machine's plans.
Since his whole purpose in inventing the Chaocipher was to allow the public a means of encryption so that anyone could send an encrypted message without failure, it seems odd that a patent application was not filed, even with only the blueprints.
In his dealings with the U.S. government, he steadfastly refused to provide anything except the cipher text and the plain text translations, after the original cigar box model was destroyed. He simply kept asking if the code he had produced was undecipherable.
I have tried to construct some models to deal with Mr. Byrne's principle, and the models which I tried to construct have no relation whatsoever to my very own CodeCircle method from JVP-Encrypt. They are different machines.
Do the models I created satisfy the contention of Mr. Byrne? Yes, in the respect that the only person to know how to decrypt the message is the person who knows the starting point. So, if everyone in the world typed in the same message, only the people who started at the same position would ever get anything similar in the result. In the models I created every message is different at a different starting point, regardless of repeating the same words over and over again.
None of my models concerning Mr. Byrne worked the way his model did. I used sets of 4 wheels and 2 wheels, and other methods, but none seemed to work in a way that produced results similar to Mr. Byrne's cigar box model.
Then, did Mr. Byrne have an undecipherable cipher?
That is a loaded question, because the answer is both YES and NO, based on preliminary mechanical examination only.
As has been stated, if we had Mr. Byrne's machine, we would know the pattern of jumbled letters. If we knew his principle, we would know how to make the cipher.
What's the secret then? Is it undecipherable code?
Mr. Byrne's cipher comes down to keeping a secret, which is the starting point in a circular sequence of chosen letters. In that case, it will remain undecipherable until someone discovers that starting point. That starting point can be quite difficult to find if you have multiple wheels with concentric disks which can be set in many ways to start the message. Mr. Byrne stated that even if someone knew the cipher and owned his machine, only the person receiving the message would know how to decipher it.
This begs an explanation, but the explanation seems to be the elusive secret starting point. If I have 4 wheels with 26 letters on each, the starting point would be the probability of the positions of all 4 wheels.
Having done a CodeCircle cipher myself, my experience in putting it together was the insight gained by ignoring the actual letters or their positions, and concentrating on the distance between the letters, in the case of CodeCircle the angular rotation of the spin. In a sense, I did not have to use a circular pattern. It could have been a slide, or I could have used distances between 2 cities in America.
The code issued by CodeCircle is nothing but angular rotation, the distance between letters on the circles. These are not the angles where the letters are located, but the spin to get from one to the other.
In other words, if you didn't know that the numbers (cipher text) were "angles," and you thought they represented "letters," you could never crack the code from CodeCircle. How long would it take you to figure out that the cipher text was a list of angles used to rotate from one letter to another?
Now, back to Mr. Byrne's Chaocipher. In a sense, as long as he doesn't reveal the starting point, his Chaocipher is indeed what he says it is, but once we know the starting point, we get the machine and run the decryption, and we decipher all of his messages.
It appears, however, that Mr. Byrne may have re-set starting points within the 100 lines of cipher text that he produced for examination. I say this, because those individuals who have examined the cipher text mathematically have found certain groups or blocks of letters similar, although nothing can be conclusive at this point. There is a difference between running the Chaocipher continuously and re-setting starting points.
I still have some of the cardboard models I constructed of Mr. Byrne's contraption, and I will probably return to them at some time in the future. I still think that, if Mr. Byrne made his method work in a cigar box model, we should be able to duplicate it mechanically, without ever doing the mathematical examination. It becomes an engineering problem at this point.
How do you make this work in a cigar box?