SG-PAC: A Stochastic Game Approach to Generate Personal Privacy Paradox Access-Control Policies in Social Networks

Yu Wu,Li Pan

DOI: https://doi.org/10.1016/j.cose.2020.102157

IF: 5.105

2020-01-01

Computers & Security

Abstract:Online social networks are indispensable platforms where people share personal data from their daily lives. However, social sharing raises many privacy issues. One of the most difficult, the “privacy paradox,” is that social-network users want to meet their social needs by enhancing interactions with their friends and are at the same time concerned about the risk of privacy leakage. This work aims to help social-network users weigh the benefits of social sharing and the costs of privacy leakage. A proposed criterion for privacy-risk measurement can quantify the probability of privacy leakage according to multi-person competitive behaviors. By analyzing the continuous competition between a user and some unwanted knowers, a general-sum stochastic game model related to the privacy paradox is constructed. The Markov decision process (MDP) and a role-mining method are introduced to facilitate the game analysis. The existence of a Nash equilibrium strategy is proved theoretically, and effective personal access-control policies are derived from this strategy as solved by a reinforcement learning algorithm. Experiments evaluate the constructed privacy stochastic game model and the new criterion for privacy-risk measurement. This work helps users develop effective personal policies in all cases to optimize their payoffs in the privacy paradox.