opendp.measurements module

opendp.measurements.make_base_discrete_gaussian(input_domain, input_metric, scale, MO='ZeroConcentratedDivergence<QO>')

Make a Measurement that adds noise from the discrete_gaussian(scale) distribution to the input.

Valid inputs for input_domain and input_metric are:

input_domain	input type	input_metric
atom_domain(T)	T	absolute_distance(QI)
vector_domain(atom_domain(T))	Vec<T>	l2_distance(QI)

make_base_discrete_gaussian in Rust documentation.

Supporting Elements:

• Input Domain: D

• Output Type: D::Carrier

• Input Metric: D::InputMetric

• Output Measure: MO

Parameters:

• input_domain – Domain of the data type to be privatized.

• input_metric – Metric of the data type to be privatized.

• scale – Noise scale parameter for the gaussian distribution. scale == standard_deviation.

• MO (Type Argument) – Output measure. The only valid measure is ZeroConcentratedDivergence<QO>, but QO can be any float.

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.make_base_discrete_laplace(input_domain, input_metric, scale, QO=None)

Make a Measurement that adds noise from the discrete_laplace(scale) distribution to the input.

Valid inputs for input_domain and input_metric are:

input_domain	input type	input_metric
atom_domain(T) (default)	T	absolute_distance(T)
vector_domain(atom_domain(T))	Vec<T>	l1_distance(T)

This uses make_base_discrete_laplace_cks20 if scale is greater than 10, otherwise it uses make_base_discrete_laplace_linear.

make_base_discrete_laplace in Rust documentation.

Citations:

• GRS12 Universally Utility-Maximizing Privacy Mechanisms

• CKS20 The Discrete Gaussian for Differential Privacy

Supporting Elements:

• Input Domain: D

• Output Type: D::Carrier

• Input Metric: D::InputMetric

• Output Measure: MaxDivergence<QO>

Parameters:

• input_domain – Domain of the data type to be privatized.

• input_metric – Metric of the data type to be privatized.

• scale – Noise scale parameter for the laplace distribution. scale == standard_deviation / sqrt(2).

• QO (Type Argument) – Data type of the output distance and scale. f32 or f64.

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.make_base_discrete_laplace_cks20(input_domain, input_metric, scale, QO=None)

Make a Measurement that adds noise from the discrete_laplace(scale) distribution to the input, using an efficient algorithm on rational bignums.

Valid inputs for input_domain and input_metric are:

input_domain	input type	input_metric
atom_domain(T) (default)	T	absolute_distance(T)
vector_domain(atom_domain(T))	Vec<T>	l1_distance(T)

make_base_discrete_laplace_cks20 in Rust documentation.

Citations:

• CKS20 The Discrete Gaussian for Differential Privacy

Supporting Elements:

• Input Domain: D

• Output Type: D::Carrier

• Input Metric: D::InputMetric

• Output Measure: MaxDivergence<QO>

Parameters:

• input_domain –

• input_metric –

• scale – Noise scale parameter for the laplace distribution. scale == standard_deviation / sqrt(2).

• QO (Type Argument) – Data type of the output distance and scale.

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.make_base_discrete_laplace_linear(input_domain, input_metric, scale, bounds=None, QO=None)

Make a Measurement that adds noise from the discrete_laplace(scale) distribution to the input, using a linear-time algorithm on finite data types.

This algorithm can be executed in constant time if bounds are passed. Valid inputs for input_domain and input_metric are:

input_domain	input type	input_metric
atom_domain(T) (default)	T	absolute_distance(T)
vector_domain(atom_domain(T))	Vec<T>	l1_distance(T)

make_base_discrete_laplace_linear in Rust documentation.

Citations:

• GRS12 Universally Utility-Maximizing Privacy Mechanisms

Supporting Elements:

• Input Domain: D

• Output Type: D::Carrier

• Input Metric: D::InputMetric

• Output Measure: MaxDivergence<QO>

Parameters:

• input_domain – Domain of the data type to be privatized.

• input_metric – Metric of the data type to be privatized.

• scale – Noise scale parameter for the distribution. scale == standard_deviation / sqrt(2).

• bounds (Any) – Set bounds on the count to make the algorithm run in constant-time.

• QO (Type Argument) – Data type of the scale and output distance.

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.make_base_gaussian(input_domain, input_metric, scale, k=-1074, MO='ZeroConcentratedDivergence<T>')

Make a Measurement that adds noise from the gaussian(scale) distribution to the input.

Valid inputs for input_domain and input_metric are:

input_domain	input type	input_metric
atom_domain(T) (default)	T	absolute_distance(T)
vector_domain(atom_domain(T))	Vec<T>	l2_distance(T)

This function takes a noise granularity in terms of 2^k. Larger granularities are more computationally efficient, but have a looser privacy map. If k is not set, k defaults to the smallest granularity.

make_base_gaussian in Rust documentation.

Supporting Elements:

• Input Domain: D

• Output Type: D::Carrier

• Input Metric: D::InputMetric

• Output Measure: MO

Parameters:

• input_domain – Domain of the data type to be privatized. Valid values are VectorDomain<AtomDomain<T>> or AtomDomain<T>.

• input_metric – Metric of the data type to be privatized. Valid values are AbsoluteDistance<T> or L2Distance<T>.

• scale – Noise scale parameter for the gaussian distribution. scale == standard_deviation.

• k (int) – The noise granularity in terms of 2^k.

• MO (Type Argument) – Output Measure. The only valid measure is ZeroConcentratedDivergence<T>.

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.make_base_geometric(input_domain, input_metric, scale, bounds=None, QO=None)

An alias for make_base_discrete_laplace_linear. If you don’t need timing side-channel protections via bounds, make_base_discrete_laplace is more efficient.

make_base_geometric in Rust documentation.

Supporting Elements:

• Input Domain: D

• Output Type: D::Carrier

• Input Metric: D::InputMetric

• Output Measure: MaxDivergence<QO>

Parameters:

• input_domain –

• input_metric –

• scale –

• bounds (Any) –

• QO (Type Argument) –

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.make_base_laplace(input_domain, input_metric, scale, k=-1074)

Make a Measurement that adds noise from the Laplace(scale) distribution to a scalar value.

Valid inputs for input_domain and input_metric are:

input_domain	input type	input_metric
atom_domain(T) (default)	T	absolute_distance(T)
vector_domain(atom_domain(T))	Vec<T>	l1_distance(T)

This function takes a noise granularity in terms of 2^k. Larger granularities are more computationally efficient, but have a looser privacy map. If k is not set, k defaults to the smallest granularity.

make_base_laplace in Rust documentation.

Supporting Elements:

• Input Domain: D

• Output Type: D::Carrier

• Input Metric: D::InputMetric

• Output Measure: MaxDivergence<D::Atom>

Parameters:

• input_domain – Domain of the data type to be privatized.

• input_metric – Metric of the data type to be privatized.

• scale – Noise scale parameter for the laplace distribution. scale == standard_deviation / sqrt(2).

• k (int) – The noise granularity in terms of 2^k.

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.make_base_laplace_threshold(input_domain, input_metric, scale, threshold, k=-1074)

Make a Measurement that uses propose-test-release to privatize a hashmap of counts.

This function takes a noise granularity in terms of 2^k. Larger granularities are more computationally efficient, but have a looser privacy map. If k is not set, k defaults to the smallest granularity.

make_base_laplace_threshold in Rust documentation.

Supporting Elements:

• Input Domain: MapDomain<AtomDomain<TK>, AtomDomain<TV>>

• Output Type: HashMap<TK, TV>

• Input Metric: L1Distance<TV>

• Output Measure: FixedSmoothedMaxDivergence<TV>

Parameters:

• input_domain – Domain of the input.

• input_metric – Metric for the input domain.

• scale – Noise scale parameter for the laplace distribution. scale == standard_deviation / sqrt(2).

• threshold – Exclude counts that are less than this minimum value.

• k (int) – The noise granularity in terms of 2^k.

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.make_gaussian(input_domain, input_metric, scale, MO='ZeroConcentratedDivergence<QO>')

Make a Measurement that adds noise from the gaussian(scale) distribution to the input.

Valid inputs for input_domain and input_metric are:

input_domain	input type	input_metric
atom_domain(T)	T	absolute_distance(QI)
vector_domain(atom_domain(T))	Vec<T>	l2_distance(QI)

make_gaussian in Rust documentation.

Supporting Elements:

• Input Domain: D

• Output Type: D::Carrier

• Input Metric: D::InputMetric

• Output Measure: MO

Parameters:

• input_domain – Domain of the data type to be privatized.

• input_metric – Metric of the data type to be privatized.

• scale – Noise scale parameter for the gaussian distribution. scale == standard_deviation.

• MO (Type Argument) – Output Measure. The only valid measure is ZeroConcentratedDivergence<T>.

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.make_laplace(input_domain, input_metric, scale, QO='float')

Make a Measurement that adds noise from the laplace(scale) distribution to the input.

Valid inputs for input_domain and input_metric are:

input_domain	input type	input_metric
atom_domain(T) (default)	T	absolute_distance(T)
vector_domain(atom_domain(T))	Vec<T>	l1_distance(T)

This uses make_base_laplace if T is float, otherwise it uses make_base_discrete_laplace.

make_laplace in Rust documentation.

Citations:

• GRS12 Universally Utility-Maximizing Privacy Mechanisms

• CKS20 The Discrete Gaussian for Differential Privacy

Supporting Elements:

• Input Domain: D

• Output Type: D::Carrier

• Input Metric: D::InputMetric

• Output Measure: MaxDivergence<QO>

Parameters:

• input_domain – Domain of the data type to be privatized.

• input_metric – Metric of the data type to be privatized.

• scale – Noise scale parameter for the laplace distribution. scale == standard_deviation / sqrt(2).

• QO (Type Argument) – Data type of the output distance and scale. f32 or f64.

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.make_randomized_response(categories, prob, constant_time=False, T=None, QO=None)

Make a Measurement that implements randomized response on a categorical value.

make_randomized_response in Rust documentation.

Supporting Elements:

• Input Domain: AtomDomain<T>

• Output Type: T

• Input Metric: DiscreteDistance

• Output Measure: MaxDivergence<QO>

Parameters:

• categories (Any) – Set of valid outcomes

• prob – Probability of returning the correct answer. Must be in [1/num_categories, 1)

• constant_time (bool) – Set to true to enable constant time. Slower.

• T (Type Argument) – Data type of a category.

• QO (Type Argument) – Data type of probability and output distance.

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.make_randomized_response_bool(prob, constant_time=False, QO=None)

Make a Measurement that implements randomized response on a boolean value.

make_randomized_response_bool in Rust documentation.

Supporting Elements:

• Input Domain: AtomDomain<bool>

• Output Type: bool

• Input Metric: DiscreteDistance

• Output Measure: MaxDivergence<QO>

Parameters:

• prob – Probability of returning the correct answer. Must be in [0.5, 1)

• constant_time (bool) – Set to true to enable constant time. Slower.

• QO (Type Argument) – Data type of probability and output distance.

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.make_user_measurement(input_domain, input_metric, output_measure, function, privacy_map, TO)

Construct a Measurement from user-defined callbacks.

make_user_measurement in Rust documentation.

Supporting Elements:

• Input Domain: AnyDomain

• Output Type: AnyObject

• Input Metric: AnyMetric

• Output Measure: AnyMeasure

Parameters:

• input_domain (Domain) – A domain describing the set of valid inputs for the function.

• input_metric (Metric) – The metric from which distances between adjacent inputs are measured.

• output_measure (Measure) – The measure from which distances between adjacent output distributions are measured.

• function – A function mapping data from input_domain to a release of type TO.

• privacy_map – A function mapping distances from input_metric to output_measure.

• TO (Type Argument) – The data type of outputs from the function.

Return type:

Measurement

Raises:

• TypeError – if an argument’s type differs from the expected type

• UnknownTypeError – if a type argument fails to parse

• OpenDPException – packaged error from the core OpenDP library

opendp.measurements.then_base_discrete_gaussian(scale, MO='ZeroConcentratedDivergence<QO>')

Parameters:

MO (RuntimeType | str | Type[List | Tuple | int | float | str | bool] | tuple | _GenericAlias | GenericAlias) –

opendp.measurements.then_base_discrete_laplace(scale, QO=None)

Parameters:

QO (RuntimeType | str | Type[List | Tuple | int | float | str | bool] | tuple | _GenericAlias | GenericAlias | None) –

opendp.measurements.then_base_discrete_laplace_cks20(scale, QO=None)

Parameters:

QO (RuntimeType | str | Type[List | Tuple | int | float | str | bool] | tuple | _GenericAlias | GenericAlias | None) –

opendp.measurements.then_base_discrete_laplace_linear(scale, bounds=None, QO=None)

Parameters:

• bounds (Any | None) –

• QO (RuntimeType | str | Type[List | Tuple | int | float | str | bool] | tuple | _GenericAlias | GenericAlias | None) –

opendp.measurements.then_base_gaussian(scale, k=-1074, MO='ZeroConcentratedDivergence<T>')

Parameters:

• k (int) –

• MO (RuntimeType | str | Type[List | Tuple | int | float | str | bool] | tuple | _GenericAlias | GenericAlias) –

opendp.measurements.then_base_geometric(scale, bounds=None, QO=None)

Parameters:

• bounds (Any | None) –

• QO (RuntimeType | str | Type[List | Tuple | int | float | str | bool] | tuple | _GenericAlias | GenericAlias | None) –

opendp.measurements.then_base_laplace(scale, k=-1074)

Parameters:

k (int) –

opendp.measurements.then_base_laplace_threshold(scale, threshold, k=-1074)

Parameters:

k (int) –

opendp.measurements.then_gaussian(scale, MO='ZeroConcentratedDivergence<QO>')

Parameters:

MO (RuntimeType | str | Type[List | Tuple | int | float | str | bool] | tuple | _GenericAlias | GenericAlias) –

opendp.measurements.then_laplace(scale, QO='float')

Parameters:

QO (RuntimeType | str | Type[List | Tuple | int | float | str | bool] | tuple | _GenericAlias | GenericAlias) –

opendp.measurements.then_user_measurement(output_measure, function, privacy_map, TO)

Parameters:

• output_measure (Measure) –

• TO (RuntimeType | str | Type[List | Tuple | int | float | str | bool] | tuple | _GenericAlias | GenericAlias) –