On the Privacy-Utility Trade-Off With and Without Direct Access to the Private Data

Abstract

We study an information theoretic privacy mechanism design problem for two scenarios where the private data is either observable or hidden. In the hidden private data scenario, an agent observes useful data Y that is correlated with private data X, and generate disclosed data U which maximizes the revealed information about Y while satisfying a bounded privacy leakage constraint. Considering the other scenario, the agent has additional access to X. To design the privacy mechanism, we first extend the Functional Representation Lemma and Strong Functional Representation Lemma by relaxing the independence condition and thereby allowing a certain leakage. We then find lower and upper bounds on the privacy-utility trade-offs in both scenarios. In particular, for the case where no leakage is allowed and X is observable, our upper and lower bounds improve previous bounds. Considering bounded mutual information as privacy constraint and the observable private data scenario we show that if the common information and mutual information between X and Y are equal, then the attained upper bound is tight. Finally, the privacy-utility trade-off with prioritized private data is studied where part of X is more private than the remaining part.

Exploring the tradeoff between data privacy and utility with a clinical data analysis use case

“…Moreover, the insights offered by such research into the tangible effects of data de-identification on actual data analysis tasks are somewhat restricted. This is because the analyses were either performed using overly simplistic examples [28,34] or on public datasets that have already undergone some form of de-identification [35,36], or focusing on theoretical aspects [37]. Therefore, there is a need for more intricate research that closely mirrors the complexities of real-life data analytics tasks and considers the multifaceted nature of data utility and privacy in actual applications.…”