Abstract
Dealing with sensitive data has been the focus of much of recent research. On one hand data disclosure may
incur some risk due to security breaches, but on the other hand data sharing has many advantages. For example,
revealing customer transactions at a grocery store may be beneficial when studying purchasing patterns and market
demand. However, a potential misuse of the revealed information may be harmful due to privacy violations. In
this paper we study the tradeoff between data disclosure and data retention. Specifically, we address the problem
of minimizing the risk of data disclosure while maintaining its utility above a certain acceptable threshold. We
formulate the problem as a discrete optimization problem and leverage the special monotonicity characteristics
for both risk and utility to construct an efficient algorithm to solve it. Such an algorithm determines the optimal
transformations that need to be performed on the microdata before it gets released. These optimal transformations
take into account both the risk associated with data disclosure and the benefit of it (referred to as utility). Through
extensive experimental studies we compare the performance of our proposed algorithm with other date disclosure
algorithms in the literature in terms of risk, utility, and time. We show that our proposed framework outperforms other techniques for sensitive data disclosure.
