local_recoding

Split records into groups and apply anonymization locally in each group.

Local recoding algorithms first attempts to divide the data into smaller groups (e.g., partitions or clusters) of at least k closely-related records. Then, anonymization is applied to records within each group, typically by assigning an identical representative value for each QID attribute. This results in several anonymized clusters of size ≥ k whose records are identical (with respect to QID); the anonymized data is simply a concatenation of these anonymized groups, thus guaranteeing k-anonymity.

Type Alias

type GroupAnonymization : (Collection[Collection], dict) -> Collection[Collection]

Prototype of a group anonymization method for local recoding algorithms.

Local recoding algorithms typically split the original records into groups. Then, a group anonymization method (implementation of this type) is applied to make records become indistinguishable in their respective group.

Parameters:

• group (Collection[Collection]) – The group of records to be anonymized.

• props (dict) – A dictionary containing necessary properties for anonymization.

Returns:

Collection[Collection] – The anonymized group.

See also

LocalRecodingAlgorithm

Abstract class for local recoding-based k-anonymization algorithms.

GroupAnonymizationBuiltIn

A set of built-in GroupAnonymization.

Classes

LocalRecodingAlgorithm	Abstract class for local recoding-based k-anonymization algorithms.
GroupAnonymizationBuiltIn	A set of built-in GroupAnonymization.
ClassicMondrian	Implementation of Classic Mondrian algorithm.
KMember	Implementation of the K-Member clustering algorithm.
OKA	Implementation of the One-Pass K-Means (OKA) clustering algorithm.