Privacy preserving through aggregating wave equation as noise in differential privacy

Abstract

The Differential Privacy (DP) is the well-known definition of privacy preserving in data mining. To preserve the privacy of the private or sensitive data, DP aggregates negligible noise to original data. The existing differential privacy uses Laplace equation as noise. The applications of DP have different attacks for Laplace noise perturbed data. This research considers time based attacks as motivation to propose a new partial differential equation i.e wave equation as noise. This research evaluates privacy performance of wave noise theoretically and practically. The experiments use numeric grids with different Ɛ coefficient as privacy coefficient. Our results indicate wave equation can also be used for privacy-preserving with Ɛ or privacy coefficient less than 1 for better knowledge extraction.