METHOD AND APPARATUS FOR UTILITY-AWARE PRIVACY PRESERVING MAPPING THROUGH ADDITIVE NOISE

摘要

that the embodiments are that the desired user is encountered to release as desired to get some utility, that is associated with the private data (indicated by S), to the analyst some public data (indicated by X) privacy - focus on the trade-offs are tailored utility. Noise privacy protection mechanism, that is to be added as Y = X + N -Y is actually release data to the analyst and N is the noise Im -, we would under 1_2- gambling distortion for the continuous data X to add Gaussian noise It shows that the optimum. We show a mechanism to add Gaussian noise to minimize the leakage of the information in the worst case by the Gaussian mechanism. Parameter for the Gaussian mechanisms are determined based on the eigenvectors of the covariance of X and specific value. We have also developed a probability privacy protection mechanism for mapping the discrete data X, where discrete random noise is up-to comply with the entropy distribution