Infer encoding mechanism from $n$-grams distance?

I have a series of obfuscated strings which across the set of strings have regions of high variance and regions of low variance – implying some encoding mechanism as opposed to an encryption procedure (or hash function) which would yield a high amount of variance across all strings; or a structure preserving cipher such as Vigenère (XOR).

I am looking for suggestions of tools and techniques for heuristics to try to identify underlying plaintext assuming this is some data encoding with English language portions (in the underlying “plaintext”).

I have considered a genetic algorithm to develop a decoding technique wherein the fitness function is “distance” of decoded text from an $n$-grams textual database. However, I would like other suggestions for solving this problem.

I have performed iterative xor brute-forcing using the tool XORtool and have bundled outputs by using the unix file command. However, this yields a high degree of false positives due to magic number matching and a large set of candidates are just “ASCII text”.

I have also used the Bletchley Tool for automatic encoding detection but have been unsatisfied with the results and general tool implementation.

• Can you get the encodings of arbitrary strings? $\;$ – user991 Oct 26 '14 at 7:34