Here’s another recent research paper that discusses automatic decryption of homophonic ciphers, and touches on the topic of the 340 not being "normal":

http://www.zodiackillerciphers.com/?p=504

I’m no cryptanalyst, but I wouldn’t put it past Z to produce the 408 which could be broken, and pair it with the 340, which is cleverly contrived gibberish. As we know, Z has/had a bizarre sense of humor and would undoubtedly enjoy the prospect of multitudes industriously attempting–and failing–to unscramble the 340.

Never once thought that the 340 was a real cipher. What better way to ensure you’re crackproof than to make up a cipher that actually says absolutely nothing?

An interesting read. It would be nice to see their source code, more specifically the CUDA portion; I have been tinkering with the idea of writing a simplified ZKD equivalent using OpenCL to use as a scoring engine for transposition candidates.

Nice idea – I’d love to see it happen!