Playing with Bacon's Cipher

Bacon's cipher is a fairly well-known and simple cipher. Its considered a form of steganography. Unlike most steganography, the code is hidden in the text and not an image. The format is a 5-bit binary encoding where each letter of the plaintext is replaced by a group of five of the letters 'A' or 'B'. This binary code has A as 00000 or "aaaaa" and Z is 11010 or "bbaba". There are variations and the original version used the same codes for the letters "I" and "J" and for "U" and "V". The usual implementation is to use one type face for "a" and a different type face for "b".

Often this is done with a serif and a sans serif font. Other options, and the first one I use below, uses italics and non-italics.

thequickbrownfoxesjumpedoverthelazy

One other option I have used is mixing upper and lower case letters.

The QUicK brOwn FoXes JUmPed ovER tHe lAZy

The big disadvantage of using mixed case letters is that it is not very subtle. It’s almost obvious that there is something special going on. Italics is less noticeable. Serif and Sans Serif are even more subtle.

One semi-famous example of the use of serif and Sans Serif fonts is the Cipher on the William and Elizebeth Friedman tombstone. The biography of Elizebeth Friedman,The Woman Who Smashed Codes: A True Story of Love, Spies, and the Unlikely Heroine Who Outwitted America's Enemies,is a seriously recommended read BTW.

I’ve been experimenting with coding and decoding using this method using a computer program. The upper/lower case combination is pretty easy both to encode and decode. C# has a nice Char.IsUpper method that was very handy for decoding. Encoding with different font faces was trickier for me. I finally used HTML for output.

Parsing an HTML file is not to bad, though I haven’t coded that at this time. The obviousness of this method still rules this out as a serious method for me. If fact just making a raw HTML file available, no matter what sort of type facing one uses, just feels to obvious for me. I suppose that getting an image of the message and posting that would work pretty well. It doesn’t lend itself to computerized decoding however. Maybe some AI pattern recognition would work well?

Speaking of images, there are examples of using this encoding in pictures of groups of people. Some people facing straight ahead and some turned to the side. Just about any possibility of two different options would work. Its a pretty flexible system.

This is not the sort of encryption that holds up to serious attack but I had some fun playing with it.

I have updated my Tiny Book of Simple Cryptography with a chapter on Bacon's Cipher. This is the seventh chapter with a simple cryptography technique.