Parallels Between Cryptanalysis and Crossword Solving

Friedman's first big development was the "index of coincidence". This is a distance measure between a crypto key being tried and the actual key. If you can tell if you're getting closer, you can hill-climb to a solution, and computers are good at this. If a distance measure can be found for a cryptosystem, it's thus easily breakable. WWII rotor machines are vulnerable to this approach. So are the classic paper-based substation cyphers, except for one-time ones where the key is as long as the text. (Use the same one-time key twice, and the cypher can be broken. See VENONA.)

Modern cryptosystem design requires that wrong decryptions look random even if the key is off by only one bit. That's a necessary, but not a sufficient, condition.