Privacy-Utility Trade-Off of K-Subset Mechanism
Yihui Zhou, Guangchen Song, Hai Liu, Laifeng Lu · 2018
In the age of big data, privacy and utility are of fundamental importance on statistical analysis. In this paper, we investigate the privacy-utility trade-off of k-subset mechanism which can be regarded as a generalization of randomized response mechanism. For a k-subset mechanism, the entropy of output random variable is described as privacy metric whereas the mutual information is considered as utility measurement. It is proved that with the increase of privacy budget, the entropy of output random variable monotonically decreases, but the mutual information increases, which illustrates the trade-off of privacy and utility.