Genetic solver for column transposition cipher

Implements a genetic algorithm solver for strings and the simple column transposition cipher. Solves a simple, singular column transposition cipher using the genetic algorithm. Probably useful if you're a spy and it's 1970.

The scoring function is the number of words (from the list of 1000 most common english words by [deekayen][github-deekayen]) present in the plaintext candidate. Words can overlap, so 'area' will be counted twice (it contains 'are'). This makes a pretty bad metric, but works for the example case and does not require spaces between words.

See __main__.py for the running example.

Usage

Proof of concept: py -m <the module>

This will find a valid key and decrypt the text in this readme correctly in about 10 seconds, most of the time.

There's no command line interface for the proof of concept, just edit __main__.py to use.

Multiplication cipher

The file cipher.py contains:

• encrypt(plaintext, key, pad=' ')
• decrypt(plaintext, key, pad=' ')

For your encryption/decription needs. A custom padding character can be specified. Using a common vowel seems to make the cipher a bit stronger against word count attacks as it will result in more 'garbage words' in the plaintext candidates.

Satisfies the example from the wiki:

>> print(encrypt('ZEBRAS', 'WEAREDISCOVEREDFLEEATONCE'))
EVLN0ACDT0ESEA0ROFO0DEEC0WIREE

String genetic-algorithm

The file ga.py contains ga_string(...) with arguments:

• score: Score function, the algorithm will try to maximise the function
• str_len: String length, default 5
• charset: Character set to choose mutations from. Defaults to uppercase ascii characters A-Z.
• gen_size: Total size of generation, default 20.
• n_generations: Number of generations, default 10
• n_survivors: Number of candidates that survive the generation. Default 5.
• mutation_rate: Number of mutated characters per mutation.

The first generation starts with gen_size fully random strings (picked from charset). Subsequent generations use the n_survivors best-fitting candidates from the previous generation, plus a number of mutations of these that brings the total number of candidates to gen_size.

Of this initial population, candidates are picked randomly and mutated such that the total population size is always n_mutations. There is no cross-over currently.

A working example is shown in main.py.