# Attacking historical ciphers methodology

It's more a theoretical question of how would you approach it. All you know about the ciphertext it's was generated with a historical cipher. The ciphertext appears to be random, BUT it's divided into sets of six letters. By historical ciphers I mean: mono (Ceasar), polyalphabetic (Vigenere), homophonic or polygraphic ciphers (Playfair, HillCipher).

I interested what methods would you use (Kasiski, Friedman, what else) and why and in what order to make the attack the most efficient (Friedman first?)

• Do you know, for example, that the plaintext is english language? Or could be plaintext be anything? – mikeazo Apr 24 '13 at 11:23
• Yes it's definitely English, only letters. – Adam Apr 24 '13 at 11:29

If you don't know the system, you just check one after the other:

• frequency analysis of bigrams detects Ceasar and Playfair. Try Caesar first then Playfair.
• Auto correlation method for Vigenere (for each x: count the number of occurances, where letter at position i and i+x are equal. For the correct codeword length, it will spike)
• If you have a Hill cipher, you lost.
• I'd do them in a slightly different order, just based on ease of the tasks: Caesar first, then Vigenere, then looking at bigrams for Playfair. I hope he has a lot of ciphertext if it's Playfair. – John Deters Apr 26 '13 at 18:51
• The Hill cipher isn't exactly unbreakable either. (Mind you, that question concerns a known-plaintext attack, but you could always do crib-dragging. Actually, it might be possible to use frequency analysis and linear algebra for a direct ciphertext-only attack... need to think about that a bit more.) – Ilmari Karonen Apr 27 '13 at 17:33
• >> If you have a Hill cipher, you lost. see this article for Hill: eprint.iacr.org/2015/802 – user26165 Aug 15 '15 at 8:20