tapas.attacks.set_classifiers.RandomTargetedQueryFeature

class tapas.attacks.set_classifiers.RandomTargetedQueryFeature(target: TabularRecord, order: int, number: int, num_bins=None)

Bases: tapas.attacks.set_classifiers.SetFeature

Features that computes random targeted queries that include the user.

These queries take the form SUM_{x in D} AND_{a in S} I{x_s = target_s}, counting all records that match the target record in all attributes in S. The set S is sampled randomly from all subsets of attributes of a fixed length (order).

This is equivalent to using entries of the order-way contingency table that count the target user as features. In particular, for order=1, this is a subset of HistSetFeature().

__init__(target: TabularRecord, order: int, number: int, num_bins=None)
Parameters

  • target (TabularRecord) – The record of the target user to attack.

  • order (int) – The number of conditions in each query.

  • number (int) – The number of queries to use.

  • num_bins (int) – How to bin continuous attributes (currently unsupported).

Methods

__init__(target, order, number[, num_bins])

param target

The record of the target user to attack.

extract(datasets)

Compute queries on each dataset.

Attributes

label

extract(datasets)

Compute queries on each dataset.