def cribDragging1(ct, keyword):
wordlist = encode_es(keyword)
ctlist = []
if len(ct) >= len(wordlist):
for i in range(len(ct) - len(wordlist)):
ct2 = []
for j, c in enumerate(wordlist):
ct2.append(ct[i + j] ^ c)
ctlist.append(ct2)
return [decode_es(ct) for ct in ctlist]
def testPos(ct, pos, wordlist):
res = []
for word in wordlist:
word_code = encode_es(word)
xor_product = [
ct[pos + i] ^ word_pos for i, word_pos in enumerate(word_code)
]
xor_product_str = decode_es(xor_product)
if not any([
x in xor_product_str
for x in ['1', '2', '3', '4', ' ', '..', ' .']
]):
prob = ngramtree.evaluateLogProb(
xor_product, ngram_length) / (len(xor_product) - ngram_length)
if not isinf(prob) and not isnan(prob) and prob < 0:
res.append((word, xor_product_str, prob))
res.sort(key=lambda tup: tup[2], reverse=True)
return res
T1 = np.zeros(shape=(stateCount, obsCount)) # store probability of the most likely path so far
T2 = np.zeros(shape=(stateCount, obsCount), dtype='uint32') # store previous most likely state in the path
# initialize matrices with first n letters
print('j: 0')
for i in range(stateCount):
T1[i,0] = unigram[charset[i]]
T2[i,0] = 0
print('j: 1')
for i in range(stateCount):
# get transition probabilities
probs = []
for k in range(stateCount):
str1 = decode_es([k, i])
str2 = decode_es([k^ct[0], i^ct[1]])
if str1 not in bigram or str2 not in bigram:
prob = -inf
else:
prob = log(T1[k, 0]) + \
log(bigram[str1]/unigram[str1[0]]) + \
log(bigram[str2]/unigram[str2[0]])
probs.append(prob)
T1[i,1] = max(probs)
if isneginf(T1[i,1]):
logger.debug('Got -inf prob at j: %d, str1: %s, str2: %s' % (1, str1, str2))
T2[i,1] = int(np.array(probs).argmax())
for j in range(2, obsCount):