find_best_caesar_break_parameters.py

   1 import random
   2 import collections
   3 from cipher import *
   4 from cipherbreak import *
   5
   6 corpus = sanitise(''.join([open('shakespeare.txt', 'r').read(),
   7     open('sherlock-holmes.txt', 'r').read(),
   8     open('war-and-peace.txt', 'r').read()]))
   9 corpus_length = len(corpus)
  10
  11 euclidean_scaled_english_counts = norms.euclidean_scale(english_counts)
  12
  13 metrics = [{'func': norms.l1, 'name': 'l1'},
  14     {'func': norms.l2, 'name': 'l2'},
  15     {'func': norms.l3, 'name': 'l2'},
  16     {'func': norms.cosine_distance, 'name': 'cosine_distance'},
  17     {'func': norms.harmonic_mean, 'name': 'harmonic_mean'},
  18     {'func': norms.geometric_mean, 'name': 'geometric_mean'},
  19     {'func': norms.inverse_log_pl, 'name': 'inverse_log_pl'}]
  20 scalings = [{'corpus_frequency': normalised_english_counts,
  21          'scaling': norms.normalise,
  22          'name': 'normalised'},
  23         {'corpus_frequency': euclidean_scaled_english_counts,
  24          'scaling': norms.euclidean_scale,
  25          'name': 'euclidean_scaled'},
  26          {'corpus_frequency': normalised_english_counts,
  27          'scaling': norms.identity_scale,
  28          'name': 'normalised_with_identity'}]
  29 message_lengths = [300, 100, 50, 30, 20, 10, 5]
  30
  31 trials = 5000
  32
  33 scores = collections.defaultdict(int)
  34
  35 with open('caesar_break_parameter_trials.csv', 'w') as f:
  36     print('metric,scaling,message_length,score', file = f)
  37     for metric in metrics:
  38         for scaling in scalings:
  39             for message_length in message_lengths:
  40                 for i in range(trials):
  41                     sample_start = random.randint(0, corpus_length - message_length)
  42                     sample = corpus[sample_start:(sample_start + message_length)]
  43                     key = random.randint(1, 25)
  44                     sample_ciphertext = caesar_encipher(sample, key)
  45                     (found_key, score) = caesar_break(sample_ciphertext,
  46                                                       metric=metric['func'],
  47                                                       target_counts=scaling['corpus_frequency'],
  48                                                       message_frequency_scaling=scaling['scaling'])
  49                     if found_key == key:
  50                         scores[(metric['name'], scaling['name'], message_length)] += 1
  51                 print(', '.join([metric['name'],
  52                                  scaling['name'],
  53                                  str(message_length),
  54                                  str(scores[(metric['name'], scaling['name'], message_length)] / trials) ]),
  55                     file = f)
  56 print()