A Better Problem Representation for k -Anonymity

Given the threat of re-identification in our growing digital society, guaranteeing privacy while providing worthwhile data for knowledge discovery has become a difficult problem. A number of heuristic solutions have been proposed that satisfy a privacy-protection property called k-Anonymity. These solutions use generalization (e.g., transform Federal Government and Local Government into a less specific occupation like Government) to ensure that each person in the database looks identical to at least k-1 other people. Without fair justification, it appears as though genetic algorithms have been discounted as viable solutions for privacy protection. This paper aims to falsify this misconception by proposing a new genetic algorithm that represents column orderings as permutations and adopts an ordered greed approach. After experimenting with real census data, we show improvements over previous genetic algorithms for privacy protection. Furthermore, in the interest of fairness, this work serves as an example of how to control experiment conditions when comparing privacy-protection algorithms.