Generalizations With Probability Distributions for Data Anonymization

Anonymization-based privacy protection ensures that data cannot be traced to an individual. To this end, an anonymizer faces two challenges. First, the output anonymization must satisfy the underlying privacy definition and second, the anonymization needs to contain as much information as possible. One way to address the latter challenge has been to introduce flexibility in value generalizations by enlarging the output domain of the algorithms. This paper presents the most flexible way of releasing generalizations by introducing the family of PDF generalizations. In a PDF generalization, each generalized data value is empowered by probability distribution functions.