LSTEG: an evolutionary game model leveraging deep reinforcement learning for privacy behavior analysis on social networks (Q6562321)

The paper presents an evolutionary game model leveraging deep reinforcement learning for privacy behavior analysis on social networks. The evolutionary game model is proposed to reveal the evolutionary processes of multi-party behaviors related to privacy as well as the parameters that determine the evolution trend. The model integrates two dynamic evolutions with different time scales characterizing the motives of three parties and their interrelationships in a network. The dynamic equations of the platform, users, and third-party companies are studied based on three defined revision protocols. The stationary-state conditions for different evolutionary terms of the game are derived. The paper also proposes a deep reinforcement learning model with adaptive evolutionary terms to solve the evolutionary stationary states in the evolutionary game model. Some experimental results are presented to reveal the complex network environment and dynamics.