Caching word segmentation
[cipher-training.git] / cipherbreak.py
1 import string
2 import collections
3 import norms
4 import logging
5 from itertools import zip_longest, cycle, permutations
6 from segment import segment
7 from multiprocessing import Pool
8 from math import log10
9
10 import matplotlib.pyplot as plt
11
12 logger = logging.getLogger(__name__)
13 logger.addHandler(logging.FileHandler('cipher.log'))
14 logger.setLevel(logging.WARNING)
15 #logger.setLevel(logging.INFO)
16 #logger.setLevel(logging.DEBUG)
17
18 from cipher import *
19 from language_models import *
20
21 # To time a run:
22 #
23 # import timeit
24 # c5a = open('2012/5a.ciphertext', 'r').read()
25 # timeit.timeit('keyword_break(c5a)', setup='gc.enable() ; from __main__ import c5a ; from cipher import keyword_break', number=1)
26 # timeit.repeat('keyword_break_mp(c5a, chunksize=500)', setup='gc.enable() ; from __main__ import c5a ; from cipher import keyword_break_mp', repeat=5, number=1)
27
28 transpositions = collections.defaultdict(list)
29 for word in keywords:
30 transpositions[transpositions_of(word)] += [word]
31
32 def frequencies(text):
33 """Count the number of occurrences of each character in text
34
35 >>> sorted(frequencies('abcdefabc').items())
36 [('a', 2), ('b', 2), ('c', 2), ('d', 1), ('e', 1), ('f', 1)]
37 >>> sorted(frequencies('the quick brown fox jumped over the lazy ' \
38 'dog').items()) # doctest: +NORMALIZE_WHITESPACE
39 [(' ', 8), ('a', 1), ('b', 1), ('c', 1), ('d', 2), ('e', 4), ('f', 1),
40 ('g', 1), ('h', 2), ('i', 1), ('j', 1), ('k', 1), ('l', 1), ('m', 1),
41 ('n', 1), ('o', 4), ('p', 1), ('q', 1), ('r', 2), ('t', 2), ('u', 2),
42 ('v', 1), ('w', 1), ('x', 1), ('y', 1), ('z', 1)]
43 >>> sorted(frequencies('The Quick BROWN fox jumped! over... the ' \
44 '(9lazy) DOG').items()) # doctest: +NORMALIZE_WHITESPACE
45 [(' ', 8), ('!', 1), ('(', 1), (')', 1), ('.', 3), ('9', 1), ('B', 1),
46 ('D', 1), ('G', 1), ('N', 1), ('O', 2), ('Q', 1), ('R', 1), ('T', 1),
47 ('W', 1), ('a', 1), ('c', 1), ('d', 1), ('e', 4), ('f', 1), ('h', 2),
48 ('i', 1), ('j', 1), ('k', 1), ('l', 1), ('m', 1), ('o', 2), ('p', 1),
49 ('r', 1), ('t', 1), ('u', 2), ('v', 1), ('x', 1), ('y', 1), ('z', 1)]
50 >>> sorted(frequencies(sanitise('The Quick BROWN fox jumped! over... ' \
51 'the (9lazy) DOG')).items()) # doctest: +NORMALIZE_WHITESPACE
52 [('a', 1), ('b', 1), ('c', 1), ('d', 2), ('e', 4), ('f', 1), ('g', 1),
53 ('h', 2), ('i', 1), ('j', 1), ('k', 1), ('l', 1), ('m', 1), ('n', 1),
54 ('o', 4), ('p', 1), ('q', 1), ('r', 2), ('t', 2), ('u', 2), ('v', 1),
55 ('w', 1), ('x', 1), ('y', 1), ('z', 1)]
56 >>> frequencies('abcdefabcdef')['x']
57 0
58 """
59 return collections.Counter(c for c in text)
60
61
62 def caesar_break(message, fitness=Pletters):
63 """Breaks a Caesar cipher using frequency analysis
64
65 >>> caesar_break('ibxcsyorsaqcheyklxivoexlevmrimwxsfiqevvmihrsasrxliwyrh' \
66 'ecjsppsamrkwleppfmergefifvmhixscsymjcsyqeoixlm') # doctest: +ELLIPSIS
67 (4, -130.849989015...)
68 >>> caesar_break('wxwmaxdgheetgwuxztgptedbgznitgwwhpguxyhkxbmhvvtlbhgtee' \
69 'raxlmhiixweblmxgxwmhmaxybkbgztgwztsxwbgmxgmert') # doctest: +ELLIPSIS
70 (19, -128.82410410...)
71 >>> caesar_break('yltbbqnqnzvguvaxurorgenafsbezqvagbnornfgsbevpnaabjurer' \
72 'svaquvzyvxrnznazlybequrvfohgriraabjtbaruraprur') # doctest: +ELLIPSIS
73 (13, -126.25403935...)
74 """
75 sanitised_message = sanitise(message)
76 best_shift = 0
77 best_fit = float('-inf')
78 for shift in range(26):
79 plaintext = caesar_decipher(sanitised_message, shift)
80 fit = fitness(plaintext)
81 logger.debug('Caesar break attempt using key {0} gives fit of {1} '
82 'and decrypt starting: {2}'.format(shift, fit, plaintext[:50]))
83 if fit > best_fit:
84 best_fit = fit
85 best_shift = shift
86 logger.info('Caesar break best fit: key {0} gives fit of {1} and '
87 'decrypt starting: {2}'.format(best_shift, best_fit,
88 caesar_decipher(sanitised_message, best_shift)[:50]))
89 return best_shift, best_fit
90
91 def affine_break(message, fitness=Pletters):
92 """Breaks an affine cipher using frequency analysis
93
94 >>> affine_break('lmyfu bkuusd dyfaxw claol psfaom jfasd snsfg jfaoe ls ' \
95 'omytd jlaxe mh jm bfmibj umis hfsul axubafkjamx. ls kffkxwsd jls ' \
96 'ofgbjmwfkiu olfmxmtmwaokttg jlsx ls kffkxwsd jlsi zg tsxwjl. jlsx ' \
97 'ls umfjsd jlsi zg hfsqysxog. ls dmmdtsd mx jls bats mh bkbsf. ls ' \
98 'bfmctsd kfmyxd jls lyj, mztanamyu xmc jm clm cku tmmeaxw kj lai ' \
99 'kxd clm ckuxj.') # doctest: +ELLIPSIS
100 ((15, 22, True), -340.601181913...)
101 """
102 sanitised_message = sanitise(message)
103 best_multiplier = 0
104 best_adder = 0
105 best_one_based = True
106 best_fit = float("-inf")
107 for one_based in [True, False]:
108 for multiplier in [x for x in range(1, 26, 2) if x != 13]:
109 for adder in range(26):
110 plaintext = affine_decipher(sanitised_message,
111 multiplier, adder, one_based)
112 fit = fitness(plaintext)
113 logger.debug('Affine break attempt using key {0}x+{1} ({2}) '
114 'gives fit of {3} and decrypt starting: {4}'.
115 format(multiplier, adder, one_based, fit,
116 plaintext[:50]))
117 if fit > best_fit:
118 best_fit = fit
119 best_multiplier = multiplier
120 best_adder = adder
121 best_one_based = one_based
122 logger.info('Affine break best fit with key {0}x+{1} ({2}) gives fit of {3} '
123 'and decrypt starting: {4}'.format(
124 best_multiplier, best_adder, best_one_based, best_fit,
125 affine_decipher(sanitised_message, best_multiplier,
126 best_adder, best_one_based)[:50]))
127 return (best_multiplier, best_adder, best_one_based), best_fit
128
129 def keyword_break(message, wordlist=keywords, fitness=Pletters):
130 """Breaks a keyword substitution cipher using a dictionary and
131 frequency analysis
132
133 >>> keyword_break(keyword_encipher('this is a test message for the ' \
134 'keyword decipherment', 'elephant', 1), \
135 wordlist=['cat', 'elephant', 'kangaroo']) # doctest: +ELLIPSIS
136 (('elephant', 1), -52.834575011...)
137 """
138 best_keyword = ''
139 best_wrap_alphabet = True
140 best_fit = float("-inf")
141 for wrap_alphabet in range(3):
142 for keyword in wordlist:
143 plaintext = keyword_decipher(message, keyword, wrap_alphabet)
144 fit = fitness(plaintext)
145 logger.debug('Keyword break attempt using key {0} (wrap={1}) '
146 'gives fit of {2} and decrypt starting: {3}'.format(
147 keyword, wrap_alphabet, fit,
148 sanitise(plaintext)[:50]))
149 if fit > best_fit:
150 best_fit = fit
151 best_keyword = keyword
152 best_wrap_alphabet = wrap_alphabet
153 logger.info('Keyword break best fit with key {0} (wrap={1}) gives fit of '
154 '{2} and decrypt starting: {3}'.format(best_keyword,
155 best_wrap_alphabet, best_fit, sanitise(
156 keyword_decipher(message, best_keyword,
157 best_wrap_alphabet))[:50]))
158 return (best_keyword, best_wrap_alphabet), best_fit
159
160 def keyword_break_mp(message, wordlist=keywords, fitness=Pletters, chunksize=500):
161 """Breaks a keyword substitution cipher using a dictionary and
162 frequency analysis
163
164 >>> keyword_break_mp(keyword_encipher('this is a test message for the ' \
165 'keyword decipherment', 'elephant', 1), \
166 wordlist=['cat', 'elephant', 'kangaroo']) # doctest: +ELLIPSIS
167 (('elephant', 1), -52.834575011...)
168 """
169 with Pool() as pool:
170 helper_args = [(message, word, wrap, fitness)
171 for word in wordlist for wrap in range(3)]
172 # Gotcha: the helper function here needs to be defined at the top level
173 # (limitation of Pool.starmap)
174 breaks = pool.starmap(keyword_break_worker, helper_args, chunksize)
175 return max(breaks, key=lambda k: k[1])
176
177 def keyword_break_worker(message, keyword, wrap_alphabet, fitness):
178 plaintext = keyword_decipher(message, keyword, wrap_alphabet)
179 fit = fitness(plaintext)
180 logger.debug('Keyword break attempt using key {0} (wrap={1}) gives fit of '
181 '{2} and decrypt starting: {3}'.format(keyword,
182 wrap_alphabet, fit, sanitise(plaintext)[:50]))
183 return (keyword, wrap_alphabet), fit
184
185 def scytale_break(message, fitness=Pbigrams):
186 """Breaks a Scytale cipher
187
188 >>> scytale_break('tfeulchtrtteehwahsdehneoifeayfsondmwpltmaoalhikotoere' \
189 'dcweatehiplwxsnhooacgorrcrcraotohsgullasenylrendaianeplscdriioto' \
190 'aek') # doctest: +ELLIPSIS
191 (6, -281.276219108...)
192 """
193 best_key = 0
194 best_fit = float("-inf")
195 for key in range(1, 20):
196 if len(message) % key == 0:
197 plaintext = scytale_decipher(message, key)
198 fit = fitness(sanitise(plaintext))
199 logger.debug('Scytale break attempt using key {0} gives fit of '
200 '{1} and decrypt starting: {2}'.format(key,
201 fit, sanitise(plaintext)[:50]))
202 if fit > best_fit:
203 best_fit = fit
204 best_key = key
205 logger.info('Scytale break best fit with key {0} gives fit of {1} and '
206 'decrypt starting: {2}'.format(best_key, best_fit,
207 sanitise(scytale_decipher(message, best_key))[:50]))
208 return best_key, best_fit
209
210
211 def column_transposition_break_mp(message, translist=transpositions,
212 fitness=Pbigrams, chunksize=500):
213 """Breaks a column transposition cipher using a dictionary and
214 n-gram frequency analysis
215 """
216 # >>> column_transposition_break_mp(column_transposition_encipher(sanitise( \
217 # "It is a truth universally acknowledged, that a single man in \
218 # possession of a good fortune, must be in want of a wife. However \
219 # little known the feelings or views of such a man may be on his \
220 # first entering a neighbourhood, this truth is so well fixed in the \
221 # minds of the surrounding families, that he is considered the \
222 # rightful property of some one or other of their daughters."), \
223 # 'encipher'), \
224 # translist={(2, 0, 5, 3, 1, 4, 6): ['encipher'], \
225 # (5, 0, 6, 1, 3, 4, 2): ['fourteen'], \
226 # (6, 1, 0, 4, 5, 3, 2): ['keyword']}) # doctest: +ELLIPSIS
227 # (((2, 0, 5, 3, 1, 4, 6), False), 0.0628106372...)
228 # >>> column_transposition_break_mp(column_transposition_encipher(sanitise( \
229 # "It is a truth universally acknowledged, that a single man in \
230 # possession of a good fortune, must be in want of a wife. However \
231 # little known the feelings or views of such a man may be on his \
232 # first entering a neighbourhood, this truth is so well fixed in the \
233 # minds of the surrounding families, that he is considered the \
234 # rightful property of some one or other of their daughters."), \
235 # 'encipher'), \
236 # translist={(2, 0, 5, 3, 1, 4, 6): ['encipher'], \
237 # (5, 0, 6, 1, 3, 4, 2): ['fourteen'], \
238 # (6, 1, 0, 4, 5, 3, 2): ['keyword']}, \
239 # target_counts=normalised_english_trigram_counts) # doctest: +ELLIPSIS
240 # (((2, 0, 5, 3, 1, 4, 6), False), 0.0592259560...)
241 # """
242 with Pool() as pool:
243 helper_args = [(message, trans, columnwise, fitness)
244 for trans in translist.keys()
245 for columnwise in [True, False]]
246 # Gotcha: the helper function here needs to be defined at the top level
247 # (limitation of Pool.starmap)
248 breaks = pool.starmap(column_transposition_break_worker,
249 helper_args, chunksize)
250 return max(breaks, key=lambda k: k[1])
251 column_transposition_break = column_transposition_break_mp
252
253 def column_transposition_break_worker(message, transposition, columnwise,
254 fitness):
255 plaintext = column_transposition_decipher(message, transposition, columnwise=columnwise)
256 fit = fitness(sanitise(plaintext))
257 logger.debug('Column transposition break attempt using key {0} '
258 'gives fit of {1} and decrypt starting: {2}'.format(
259 transposition, fit,
260 sanitise(plaintext)[:50]))
261 return (transposition, columnwise), fit
262
263
264 def transposition_break_exhaustive(message, fitness=Pbigrams):
265 best_transposition = ''
266 best_pw = float('-inf')
267 for keylength in range(1, 21):
268 if len(message) % keylength == 0:
269 for transposition in permutations(range(keylength)):
270 for columnwise in [True, False]:
271 plaintext = column_transposition_decipher(message,
272 transposition, columnwise=columnwise)
273 fit=fitness(plaintext)
274 logger.debug('Column transposition break attempt using key {0} {1} '
275 'gives fit of {2} and decrypt starting: {3}'.format(
276 transposition, columnwise, pw,
277 sanitise(plaintext)[:50]))
278 if fit > best_fit:
279 best_transposition = transposition
280 best_columnwise = columnwise
281 best_fit = fit
282 return (best_transposition, best_columnwise), best_pw
283
284
285 def vigenere_keyword_break(message, wordlist=keywords, fitness=Pletters):
286 """Breaks a vigenere cipher using a dictionary and
287 frequency analysis
288
289 >>> vigenere_keyword_break(vigenere_encipher(sanitise('this is a test ' \
290 'message for the vigenere decipherment'), 'cat'), \
291 wordlist=['cat', 'elephant', 'kangaroo']) # doctest: +ELLIPSIS
292 ('cat', -52.947271216...)
293 """
294 best_keyword = ''
295 best_fit = float("-inf")
296 for keyword in wordlist:
297 plaintext = vigenere_decipher(message, keyword)
298 fit = fitness(plaintext)
299 logger.debug('Vigenere break attempt using key {0} '
300 'gives fit of {1} and decrypt starting: {2}'.format(
301 keyword, fit,
302 sanitise(plaintext)[:50]))
303 if fit > best_fit:
304 best_fit = fit
305 best_keyword = keyword
306 logger.info('Vigenere break best fit with key {0} gives fit '
307 'of {1} and decrypt starting: {2}'.format(best_keyword,
308 best_fit, sanitise(
309 vigenere_decipher(message, best_keyword))[:50]))
310 return best_keyword, best_fit
311
312 def vigenere_keyword_break_mp(message, wordlist=keywords, fitness=Pletters,
313 chunksize=500):
314 """Breaks a vigenere cipher using a dictionary and
315 frequency analysis
316
317 >>> vigenere_keyword_break_mp(vigenere_encipher(sanitise('this is a test ' \
318 'message for the vigenere decipherment'), 'cat'), \
319 wordlist=['cat', 'elephant', 'kangaroo']) # doctest: +ELLIPSIS
320 ('cat', -52.947271216...)
321 """
322 with Pool() as pool:
323 helper_args = [(message, word, fitness)
324 for word in wordlist]
325 # Gotcha: the helper function here needs to be defined at the top level
326 # (limitation of Pool.starmap)
327 breaks = pool.starmap(vigenere_keyword_break_worker, helper_args, chunksize)
328 return max(breaks, key=lambda k: k[1])
329
330 def vigenere_keyword_break_worker(message, keyword, fitness):
331 plaintext = vigenere_decipher(message, keyword)
332 fit = fitness(plaintext)
333 logger.debug('Vigenere keyword break attempt using key {0} gives fit of '
334 '{1} and decrypt starting: {2}'.format(keyword,
335 fit, sanitise(plaintext)[:50]))
336 return keyword, fit
337
338
339
340 def vigenere_frequency_break(message, fitness=Pletters):
341 """Breaks a Vigenere cipher with frequency analysis
342
343 >>> vigenere_frequency_break(vigenere_encipher(sanitise("It is time to " \
344 "run. She is ready and so am I. I stole Daniel's pocketbook this " \
345 "afternoon when he left his jacket hanging on the easel in the " \
346 "attic. I jump every time I hear a footstep on the stairs, " \
347 "certain that the theft has been discovered and that I will " \
348 "be caught. The SS officer visits less often now that he is " \
349 "sure"), 'florence')) # doctest: +ELLIPSIS
350 ('florence', -307.5473096791...)
351 """
352 best_fit = float("-inf")
353 best_key = ''
354 sanitised_message = sanitise(message)
355 for trial_length in range(1, 20):
356 splits = every_nth(sanitised_message, trial_length)
357 key = ''.join([chr(caesar_break(s)[0] + ord('a')) for s in splits])
358 plaintext = vigenere_decipher(sanitised_message, key)
359 fit = fitness(plaintext)
360 logger.debug('Vigenere key length of {0} ({1}) gives fit of {2}'.
361 format(trial_length, key, fit))
362 if fit > best_fit:
363 best_fit = fit
364 best_key = key
365 logger.info('Vigenere break best fit with key {0} gives fit '
366 'of {1} and decrypt starting: {2}'.format(best_key,
367 best_fit, sanitise(
368 vigenere_decipher(message, best_key))[:50]))
369 return best_key, best_fit
370
371 def beaufort_frequency_break(message, fitness=Pletters):
372 """Breaks a Beaufort cipher with frequency analysis
373
374 >>> beaufort_frequency_break(beaufort_encipher(sanitise("It is time to " \
375 "run. She is ready and so am I. I stole Daniel's pocketbook this " \
376 "afternoon when he left his jacket hanging on the easel in the " \
377 "attic. I jump every time I hear a footstep on the stairs, " \
378 "certain that the theft has been discovered and that I will " \
379 "be caught. The SS officer visits less often now " \
380 "that he is sure"), 'florence')) # doctest: +ELLIPSIS
381 ('florence', -307.5473096791...)
382 """
383 best_fit = float("-inf")
384 best_key = ''
385 sanitised_message = sanitise(message)
386 for trial_length in range(1, 20):
387 splits = every_nth(sanitised_message, trial_length)
388 key = ''.join([chr(-caesar_break(s)[0] % 26 + ord('a')) for s in splits])
389 plaintext = beaufort_decipher(sanitised_message, key)
390 fit = fitness(plaintext)
391 logger.debug('Beaufort key length of {0} ({1}) gives fit of {2}'.
392 format(trial_length, key, fit))
393 if fit > best_fit:
394 best_fit = fit
395 best_key = key
396 logger.info('Beaufort break best fit with key {0} gives fit '
397 'of {1} and decrypt starting: {2}'.format(best_key,
398 best_fit, sanitise(
399 beaufort_decipher(message, best_key))[:50]))
400 return best_key, best_fit
401
402
403
404 def plot_frequency_histogram(freqs, sort_key=None):
405 x = range(len(freqs.keys()))
406 y = [freqs[l] for l in sorted(freqs.keys(), key=sort_key)]
407 f = plt.figure()
408 ax = f.add_axes([0.1, 0.1, 0.9, 0.9])
409 ax.bar(x, y, align='center')
410 ax.set_xticks(x)
411 ax.set_xticklabels(sorted(freqs.keys(), key=sort_key))
412 f.show()
413
414
415 if __name__ == "__main__":
416 import doctest
417 doctest.testmod()
418