Privacy-Preserving Correlated Data Publication: Privacy Analysis and Optimal Noise Design

The privacy issue in data publication is critical and has been extensively studied. Correlation unavoidable publication, which universally manifests intrinsic correlations owing to social, physical, behavioral, genetic relationships. However, most of the existing works assume that private independent, i.e., correlation among neglected. In this paper, we investigate concern where deterministic probabilistic are considered, respectively. Specifically, $(\varepsilon, \delta)$ -multi-dimensional data-privacy (MDDP) proposed quantify correlated privacy. It characterizes disclosure probability published being jointly estimated with under a given accuracy. Then, explore effects on disclosure, For both kinds correlations, it shown increases compared one without knowledge. Meanwhile, closed-form expression strict bound gain derived, To minimize probability, provide optimal noise distribution sense -MDDP. Extensive simulations real dataset verify our analytical results.