wolpert@hpislx.HP.COM (David Wolpert) (04/26/88)
Statistical Tests
I'm interested in finding out what statistical tests could be used
to decide what method of encryption was used on a text. I'm assuming
"English" or similar languages, here, and simple encryption schemes
(substitution, transposition, digrams, etc). I looked in Kahn
(_The_Codebreakers_) and found the "Kappa" tests, to decide if a
cryptogram was mono- or poly-alphabetic, that seemed as if it would
be helpful for some stuff. Are there similar tests for digram or
matrix ciphers (or do I have to derive the constants myself from
digram tables?).
I looked in Gaines (_Elementary_Cryptanalysis_), and the main test
to decide what cipher method was used (after frequency counting of
letters and digrams) seems to be "try it". I'm interested in some
formula, I can program for a small computer, to examine data files
and print out something like "file CIPHER.D1 is probably monalphabetic
substitution" or "file CIPHER.D6 is probably a variant of Playfair".
Or does this break down too quickly as the ciphers get better
(ciphertext more random)? Thanks for any help; please mail. If you
want me to post results, ask by mail.
_ _
/ \/ \ David Wolpert
/_ __ HEWLETT \ / Measurement Systems Operation
/ / /_/ PACKARD \ / P O Box 301 - Loveland, CO 80539
/ \/ wolpert%hpisla@hplabs.HP.COM
======================================================================
"Be joyful always; pray continually; give thanks in all circumstances"