def test_discrete_laplace_privacy_loss(self, parameter, sensitivity, x,
expected_privacy_loss):
pld = privacy_loss_distribution.DiscreteLaplacePrivacyLossDistribution(
parameter,
sensitivity=sensitivity,
value_discretization_interval=1)
self.assertAlmostEqual(expected_privacy_loss, pld.privacy_loss(x))
def test_laplace_get_delta_for_epsilon(self, parameter, sensitivity,
epsilon, expected_divergence):
pld = privacy_loss_distribution.DiscreteLaplacePrivacyLossDistribution(
parameter,
sensitivity=sensitivity,
value_discretization_interval=1)
self.assertAlmostEqual(expected_divergence,
pld.get_delta_for_epsilon(epsilon))
def test_discrete_laplace_privacy_loss_value_errors(
self, parameter, sensitivity, x):
pld = privacy_loss_distribution.DiscreteLaplacePrivacyLossDistribution(
parameter,
sensitivity=sensitivity,
value_discretization_interval=1)
with self.assertRaises(ValueError):
pld.privacy_loss(x)
def test_discrete_laplace_varying_standard_deviation_and_sensitivity(
self, parameter, sensitivity, expected_rounded_probability_mass_function):
pld = privacy_loss_distribution.DiscreteLaplacePrivacyLossDistribution(
parameter, sensitivity=sensitivity, value_discretization_interval=1)
dictionary_almost_equal(
self,
expected_rounded_probability_mass_function,
pld.rounded_probability_mass_function)
def test_discrete_laplace_optimistic(
self, parameter, sensitivity, expected_rounded_probability_mass_function):
pld = privacy_loss_distribution.DiscreteLaplacePrivacyLossDistribution(
parameter, sensitivity=sensitivity,
pessimistic_estimate=False, value_discretization_interval=1)
dictionary_almost_equal(
self,
expected_rounded_probability_mass_function,
pld.rounded_probability_mass_function)
def test_discrete_laplace_discretization(
self, value_discretization_interval,
expected_rounded_probability_mass_function):
# Set parameter = 1, sensitivity = 2 here.
pld = privacy_loss_distribution.DiscreteLaplacePrivacyLossDistribution(
1,
sensitivity=2,
value_discretization_interval=value_discretization_interval)
dictionary_almost_equal(self,
expected_rounded_probability_mass_function,
pld.rounded_probability_mass_function)
def test_discrete_laplace_privacy_loss_tail(
self, parameter, sensitivity, expected_lower_x_truncation,
expected_upper_x_truncation, expected_tail_probability_mass_function):
pld = privacy_loss_distribution.DiscreteLaplacePrivacyLossDistribution(
parameter, sensitivity=sensitivity, value_discretization_interval=1)
tail_pld = pld.privacy_loss_tail()
self.assertAlmostEqual(expected_lower_x_truncation,
tail_pld.lower_x_truncation)
self.assertAlmostEqual(expected_upper_x_truncation,
tail_pld.upper_x_truncation)
dictionary_almost_equal(self, expected_tail_probability_mass_function,
tail_pld.tail_probability_mass_function)
def test_discrete_laplace_value_errors(self, parameter, sensitivity):
with self.assertRaises(ValueError):
privacy_loss_distribution.DiscreteLaplacePrivacyLossDistribution(
parameter, sensitivity=sensitivity)
