def predict(self, key_data):
ref_key_attributes = closest_neighbors(self.synthetic_dict.keys(),
key_data)
ref_sensitive_attributes = []
for key in ref_key_attributes:
ref_sensitive_attributes.extend(self.synthetic_dict[key])
return majority(ref_sensitive_attributes)
def score(self, key_data, sensitive_data):
ref_key_attributes = closest_neighbors(self.synthetic_dict.keys(),
key_data)
ref_sensitive_attributes = []
for key in ref_key_attributes:
ref_sensitive_attributes.extend(self.synthetic_dict[key])
return count_frequency(ref_sensitive_attributes, sensitive_data)
def test_closest_neighbors_exact():
samples = [
('a', '1'),
('a', '2'),
('a', '3'),
('b', '1'),
('b', '2'),
('b', '3'),
]
target = ('a', '2')
results = closest_neighbors(samples, target)
assert len(results) == 1
assert results[0] == ('a', '2')
def test_closest_neighbors_non_exact():
samples = [
('a', '1'),
('a', '3'),
('b', '1'),
('b', '2'),
('b', '3'),
]
target = ('a', '2')
results = closest_neighbors(samples, target)
assert len(results) == 3
assert ('a', '1') in results
assert ('a', '3') in results
assert ('b', '2') in results
def predict(self, key_data):
"""Make a prediction of the sensitive data given keys.
Args:
key_data (tuple):
The key data.
Returns:
tuple:
The predicted sensitive data.
"""
ref_key_attributes = closest_neighbors(self.synthetic_dict.keys(), key_data)
ref_sensitive_attributes = []
for key in ref_key_attributes:
ref_sensitive_attributes.extend(self.synthetic_dict[key])
return majority(ref_sensitive_attributes)
def score(self, key_data, sensitive_data):
"""Score based on the belief of the attacker, in the form P(sensitive_data|key|data).
Args:
key_data (tuple):
The key data.
sensitive_data (tuple):
The sensitive data.
Returns:
float or None:
The frequency of the correct sensitive entry.
"""
ref_key_attributes = closest_neighbors(self.synthetic_dict.keys(), key_data)
ref_sensitive_attributes = []
for key in ref_key_attributes:
ref_sensitive_attributes.extend(self.synthetic_dict[key])
return count_frequency(ref_sensitive_attributes, sensitive_data)