Evaluating applicability of perturbation techniques for privacy preserving data mining by descriptive statistics

Extensive research has been carried out for preserving the privacy of identifiers in dataset during Data Mining. Various dimensions based on Cryptographic principles, Perturbation and Secure Sum Computation have been studied to achieve privacy. Effective techniques to maximize privacy and minimize information loss have always been intriguing. The work in this paper presents a comparison based on experimental study of three fundamental perturbation techniques viz. — Additive, Multiplicative and Geometric Data Perturbation [GDP] for Privacy Preserving Data Mining [PPDM]. These techniques form the basis of many advanced Perturbation techniques as described later. The literature doesn't embark a clear cut comparison amongst the three techniques based on suitable metrics. We have identified various statistical metrics that must be considered for evaluating Perturbation techniques. The facet of research is independent in this context, and this paper will try to confer the applicability of perturbation techniques by descriptive statistics through experiments under one roof. A comparison amongst the perturbation based techniques is conferred at the end to exemplify the importance of this research.