kaplan_meier.data_owner module

Module implements the data owners Alice and Bob.

class kaplan_meier.data_owner.Alice(*args, time_label='time', event_label='event', security_level=40, **kwargs)

Bases: DataOwner

Alice player in the MPC protocol

__init__(*args, time_label='time', event_label='event', security_level=40, **kwargs)

Initializes player Alice

Parameters:

• time_label (str) – the label used to represent the ‘time’ column in the data set

• event_label (str) – the label used to represent the ‘event’ column in the data set

• security_level (int | None) – the statistical security level to use for the additive masking (in bits)

• args (Any) – arguments to pass on to base class

• kwargs (Any) – keyword arguments to pass on to base class

compute_kaplan_meier_minimal_statistics()

Method to compute the minimal statistics of the protocol.

Return type:

None

compute_logrank_factors()

Pre-computes several factors for in the computation of the log- rank statistic, leveraging information known by Alice only.

Computes the following factors: expectation_factors, variance_factors, variance_factors_2. These factors satisfy the following relations: :rtype: None

Expected number of observations in group i =

expectation_factors[i] * at_risk_group[i]

Variance of observations in group i =

(variance_factors_2[i] - variance_factors[i] * at_risk_group[i]) * at_risk_group[i]

generate_additive_share()

Generates additive secret shares.

Return type:

None

property kaplan_meier_statistics_group: ndarray[Any, dtype[object_]]

Encrypted table with Kaplan-Meier statistics per group.

Row i contains data for the i-th unique event times. The first set of (groups-1) columns contains the number of observed events for group 1, 2, …, groups-1. The second set of (groups-1) columns contains the number of individuals at risk for those groups.

Returns:

the constructed table

Raises:

ValueError – raised when table is not yet available

property kaplan_meier_statistics_total: ndarray[Any, dtype[int64]]

Table with Kaplan-Meier statistics of the entire population.

Row i contains data for the i-th unique event times. The first column contains the number of observed events for the entire population. The second column contains the number of individuals at risk for the entire population.

Returns:

the constructed table

Raises:

ValueError – raised when table is not yet available

mask_kaplan_meier_statistics_group()

Masks the table with Kaplan-meier group statistics.

Return type:

ndarray[Any, dtype[object_]]

Returns:

the masked table

property nr_groups: int

Number of groups in the datasets.

Returns:

number of groups

Raises:

ValueError – raised when number of groups is not available (yet)

property nr_unique_event_times: int

Number of unique event times.

Returns:

number of unique event times

async receive_encrypted_group_data()

Method to receive the encrypted group data from party Bob.

Return type:

None

async receive_number_of_groups()

Method to receive the number of groups identified by party Bob.

Return type:

None

async receive_paillier_scheme()

Method to receive the Paillier scheme that is used by party Bob.

Return type:

None

async run_protocol()

Starts and runs the protocol

Return type:

None

async send_additive_share()

Sends additive secret share to party Bob.

Return type:

None

signed_randomness()

Returns a signed random plaintext value.

Return type:

SupportsInt

Returns:

signed random plaintext value

property unique_event_times: ndarray[Any, dtype[int64]]

List of unique event times.

Returns:

unique event times

class kaplan_meier.data_owner.Bob(*args, paillier_scheme=<tno.mpc.encryption_schemes.paillier.paillier.Paillier object>, **kwargs)

Bases: DataOwner

Bob player in the MPC protocol

__init__(*args, paillier_scheme=<tno.mpc.encryption_schemes.paillier.paillier.Paillier object>, **kwargs)

Initializes player Bob

Parameters:

• paillier_scheme (Paillier) – the Paillier scheme to use for encryption

• args (Any) – arguments to pass on to base class

• kwargs (Any) – keyword arguments to pass on to base class

decrypt(data)

Method to decrypt a dataset using the initialized Paillier scheme

Parameters:

data (ndarray[Any, dtype[object_]]) – the dataset to decrypt

Return type:

ndarray[Any, dtype[int64]]

Returns:

a decrypted dataset

async decrypt_additive_share(data)

Decrypt share

Parameters:

data (ndarray[Any, dtype[object_]]) – the dataset (share) to decrypt

Return type:

ndarray[Any, dtype[int64]]

Returns:

decrypted data set

property nr_groups: int

Number of groups in the loaded dataset

Returns:

number of groups

async receive_additive_share()

Receive additive secret share produced by party Alice.

Return type:

None

async run_protocol()

Starts and runs the protocol

Return type:

None

async send_encrypted_group_data()

Sends the encrypted dataset to party Alice.

Return type:

None

async send_number_of_groups()

Sends the number of groups to party Alice.

Return type:

None

async send_paillier_scheme()

Sends the used Paillier scheme to party Alice.

Return type:

None

class kaplan_meier.data_owner.DataOwner(data, pool, *args, **kwargs)

Bases: Player

Data owner in the MPC protocol

__init__(data, pool, *args, **kwargs)

Initializes data owner

Parameters:

• data (DataFrame) – the data to use for this data owner

• pool (Pool) – a communication pool

• *args (Any) – arguments to pass on to base class

• **kwargs (Any) – keyword arguments to pass on to base class

property data: DataFrame

The loaded dataset

Returns:

dataset

Raises:

ValueError – raised when there is no data available

encrypt(data)

Method to encrypt a dataset using the initialized Paillier scheme

Parameters:

data (ndarray[Any, dtype[int64]]) – the dataset to encrypt

Return type:

ndarray[Any, dtype[object_]]

Returns:

an encrypted dataset

property nr_groups: int

Number of groups in the loaded datasets

Returns:

number of groups

Raises:

NotImplementedError – raised when not implemented

property paillier_scheme: Paillier

The Paillier scheme

Returns:

Paillier scheme

Raises:

ValueError – raised when Paillier scheme is not available yet.

async receive_message(party, msg_id=None)

Receives a message from a party (belonging to an optional message identifier)

Parameters:

• party (str) – the party to receive a message from

• msg_id (str | None) – the message id

Return type:

Any

Returns:

the received message

property records: int

Number of records in the loaded dataset

Returns:

number of records

async send_message(receiver, message, msg_id=None)

Sends a message to a party (with an optional message identifier)

Parameters:

• receiver (str) – the party to send a message to

• message (Any) – the message to send

• msg_id (str | None) – the message id

Return type:

None

shut_down_paillier()

Shut down Paillier scheme.

Return type:

None

start_randomness_generation(amount)

Kicks off the randomness generation. This boosts performance. In particular will this decrease the total runtime (as data owners can already generate randomness before they need it).

Parameters:

amount (int) – amount of randomness to precompute.

Return type:

None

kaplan_meier.data_owner.randomize_ndarray(arr)

Randomize all elements in an numpy array with ciphertexts.

This function calls ‘RandomizableCiphertext.randomize’ ‘arr.size’ times, as expected. Note that this contrasts with the ‘arr.size+1’ calls made by ‘np.vectorize(lambda _: _.randomize())(arr)’.

Parameters:

arr (ndarray[Any, dtype[object_]]) – array to be randomized

Return type:

None