Apparatus and methods for anonymizing a data set

A non-transitory computer readable storage medium storing program instructions executable by one or more processors on one or more computers to determine clusters of similar entities from a set of entities by:

• partitioning the set of entities into a first group of entities with similar attributes to one another and a complement group of entities with similar attributes to one another, wherein the program instructions cause the computer to perform the partitioning by:
• choosing a reference entity from the set of entities;
• determining a symmetric set of attributes based on the reference entity's attributes and on an average of the attributes of the set of entities being partitioned, wherein the average of the attributes of the set of entities being partitioned is calculated as a centroid of the attributes, and wherein the symmetric set of attributes of a reference entity is determined as a reflection of attributes of the reference entity about the centroid;
• and
• assigning each entity to the first group or to the complement group depending on whether the entity's attributes are more similar to those of the reference entity or to those of the symmetric set of attributes respectively;
• identifying the clusters of similar entities by recursively repeating the partitioning on the first group and on the complement group until every group meets one or more size criteria;
• and
• creating an anonymous version of the dataset, wherein creating the anonymous version comprises:
• calculating, for one of the clusters of similar entities, an anonymous value for a given attribute;
• and
• assigning the anonymous value of the attribute to each entity in the cluster.

An apparatus comprising:

• one or more processors,
• memory coupled to the one or more processors, wherein the memory stores program instructions executable by the one or more processors to determine clusters of similar entities from a set of entities by:
• partitioning the set of entities into a first group of entities with similar attributes to one another and a complement group of entities with similar attributes to one another, wherein the instructions, when executed by the one or more processors, causes the processor to perform the partitioning by:
• choosing a reference entity from the set of entities;
• determining a symmetric set of attributes based on the reference entity's attributes and on an average of the attributes of the set of entities being partitioned, wherein the average of the attributes of the set of entities being partitioned is calculated as a centroid of the attributes, and wherein the symmetric set of attributes of a reference entity is determined as a reflection of attributes of the reference entity about the centroid;
• and
• assigning each entity to the first group or to the complement group depending on whether the entity's attributes are more similar to those of the reference entity or to those of the symmetric set of attributes respectively;
• identifying the clusters of similar entities by recursively repeating the partitioning on the first group and on the complement group until every group meets one or more criteria;
• and
• creating an anonymous version of the dataset, wherein creating the anonymous version comprises:
• calculating, for one of the clusters of similar entities, an anonymous value for a given attribute;
• and
• assigning the anonymous value of the attribute to each entity in the cluster.