Abstract
Differential privacy schemes have been widely adopted in recent years to address issues of data privacy protection. We propose a new Gaussian scheme combining with another data protection technique, called random orthogonal matrix masking, to achieve (ε, δ)-differential privacy (DP) more efficiently. We prove that the additional matrix masking significantly reduces the rate of noise variance required in the Gaussian scheme to achieve (ε, δ)-DP in big data setting. Specifically, when ε → 0, δ → 0, and the sample size n exceeds the number p of attributes by (n-p)=O(ln(1/δ)), the required additive noise variance to achieve (ε, δ)-DP is reduced from O(ln(1/δ)/ε²) to O(1/ε). With much less noise added, the resulting differential privacy protected pseudo data sets allow much more accurate inferences, thus can significantly improve the scope of application for differential privacy.