def pair_comparison(subjects_data, tasks_list):
'''Scenario 0: Get the highest mean cross_validation score of two taks, from all possible task pairs within subject.'''
log_file_name = 'Scenario0_log.txt'
log = open(log_file_name, 'w')
print("Scenario 0: Pair Comparison")
for target_subject in sorted(subjects_data.keys()):
print('For subject ' + str(target_subject) + ' : ')
logger = '{"' + target_subject + '":{'
all_possible_pairs = combinations(tasks_list, 2)
scores_holder = []
for tasks_pair in all_possible_pairs:
task1 = reduce(lambda x, y: x + y, subjects_data[target_subject][0][tasks_pair[0]][0])
# print(len(task1))
# print(len(task1[0]), len(task1[1]), len(task1[2]), len(task1[3]), len(task1[4]))
task2 = reduce(lambda x, y: x + y, subjects_data[target_subject][0][tasks_pair[1]][0])
# print(len(task2))
X, y = learner.labeler([task1, task2])
mean, std = learner.crossValidate(X, y)
scores_holder.append([mean, std, tasks_pair[0], tasks_pair[1]])
# print(tasks_pair[0], tasks_pair[1], mean, std)
logger += '"' + tasks_pair[0] + ' vs. ' + tasks_pair[1] + '": "' + str(scores_holder[-1][0]) + '", '
best_pair = max(scores_holder, key=itemgetter(0))
print(best_pair)
logger = logger[:-2] + '}}\n'
log.write(logger)
log.close()
return
def passthought_leakage(subjects_data, tasks_list, times):
'''Scenario 2: A subject's passthought was known and attackers try to impersonate him.'''
log_file_name = 'Scenario2_log.txt'
log = open(log_file_name, 'w')
print("Scenario 2: Passthought Leakage")
for target_subject in sorted(subjects_data.keys()):
print('For subject' + str(target_subject) + ': ')
logger = '{"' + target_subject + '":{'
subject_sample_numbers = subjects_data[target_subject][1]
for task in tasks_list:
print('For task: ' + task)
logger += '"' + task + '":{'
task_sample_numbers = sum(subjects_data[target_subject][0][task][1])
others_list = []
others_subjects_sample_numbers = []
others_sample_numbers = []
for other_subject in sorted(subjects_data.keys()):
if other_subject == target_subject:
continue
else:
sample_numbers = sum(subjects_data[other_subject][0][task][1])
others_list.append(other_subject)
others_subjects_sample_numbers.append(sample_numbers)
others_sample_numbers.append([sample_numbers])
FRR_holder = []
FAR_holder = []
for one_sample_run in range(times):
logger += '"' + str(one_sample_run) + '":{'
trainer, tester = selector.random_sampler(task_sample_numbers)
logger += '"target":"' + str(trainer) + ',' + str(tester)
train_data_target = selector.sample_mapper(trainer, [target_subject], [task], [subject_sample_numbers], [[task_sample_numbers]], subjects_data)
test_data_target = selector.sample_mapper(tester, [target_subject], [task], [subject_sample_numbers], [[task_sample_numbers]], subjects_data)
trainer, tester = selector.random_sampler(sum(others_subjects_sample_numbers))
logger += '","other":"' + str(trainer) + ',' + str(tester) + '"}, '
train_data_other = selector.sample_mapper(trainer, others_list, [task], others_subjects_sample_numbers, others_sample_numbers, subjects_data)
test_data_other = selector.sample_mapper(tester, others_list, [task], others_subjects_sample_numbers, others_sample_numbers, subjects_data)
X, y = learner.labeler([train_data_target, train_data_other])
lin_clf = svm.LinearSVC()
lin_clf.fit(X, y)
test_result_correct = lin_clf.predict(test_data_target)
FRR_holder.append(np.mean(test_result_correct))
test_result_wrong = lin_clf.predict(test_data_other)
FAR_holder.append(1 - np.mean(test_result_wrong))
logger = logger[:-2] + '}, '
print(' FRR = ' + str(np.mean(FRR_holder)) + ', FAR = ' + str(np.mean(FAR_holder)))
logger = logger[:-2] + '}}\n'
log.write(logger)
log.close()
return
