All this talk of stream ciphers and masking techniques has gotten me wondering.

Is it possible that Ed never meant for us to need clues to solve Kryptos?  I would assume that he and Sanborn talked about leaving hints and clues throughout but I think Ed was smart enough to consider that some wouldn’t need them or find them or understand them.

I’m not saying there aren’t clues, just considering what they would have considered when planning.

So if our projected cryptanalyst approaches Kryptos he will rapidly translate the Morse code.  He will then proceed into the courtyard and upon seeing the vigenere side and the ciphertext, quickly get to work cracking the first part, then the second and the third.  For someone breaking the code, it quite often doesn’t matter much what it actually says so it’s possible that a trained analyst could defeat Kryptos up to this point.  S(he) would get to K4 and probably assume a cipher that masks letter frequencies.

Here’s where we take a time-out and think about things.

If there are clues that are too explicit then Kryptos was a joke that anyone could figure out.  Anyone who’s done some background work will get the sense that it was meant for entertainment but not as a joke.  So we get some clues that hint around at things, but nothing too specific. 

Let’s go back to our analyst, Ed would know that just a masking method would be too easy so he ciphered Sanborn’s text and then masked it.  Solve the masking and then you still have the ciphertext.  Put yourself in his shoes, would you make the cipher so ridiculously hard that it would take a lot of time to solve it.  No way.  Why?  If you make the ciphertext too hard then an analyst would never know if they’d solved the masking method or not.

“Ahhh”, you interject, “but the clues!”. 

If, and this is a big if, Ed planned it out for a typical code-, cipher-breaking analyst he would not have depended upon the context of the plaintext to help us solve the rest.  This is why I’d make the argument that the underlying ciphertext will be identifiable in some way as to let us know either by frequency analysis or by some other means that we’ve made it through.

Without this assumption it would be fairly easy to decipher with numerous sequential methods until you’re convinced you can finally crack it with just one more…

How would you ever know when you’ve broken through and have obtained the ciphertext?

It’s an argument that smacks of teleology but right or wrong, it’s worth a couple minutes to consider it.