Risk & Distortion Based K-Anonymity.

Current optimizations for K-Anonymity pursue reduction of data distortion unilaterally, and rarely evaluate disclosure risk during process of anonymization. We propose an optimal K-Anonymity algorithm in which the balance of risk & distortion $\left(RD\right)$ can be equilibrated at each anonymity stage: we first construct a generalization space $\left(GS\right)$, then, we use the probability and entropy metric to measure RD for each node in GS, and finally we introduce releaser's RD preference to decide an optimal anonymity path. Our algorithm adequately considers the dual-impact on RD and obtains an optimal anonymity with satisfaction of releaser. The efficiency of our algorithm will be evaluated by extensive experiments. [ABSTRACT FROM AUTHOR]