Non-Uniform Influence Blocking Maximization in Social Network

Online Social Network (OSN) is one of the most popular internet services. It also has become the main source of news for many people. Despite all the benefits, OSN significantly increases the rate of rumor spreading among people. Influence Blocking Maximization (IBM) aims to limit the propagation of rumor by broadcasting anti-rumor information. In IBM problem, the users are treated equally, however, they may have dissimilar worthiness. In this paper, we introduce Non-Uniform IBM (NU-IBM) where each user has its own weight. As a case study for NU-IBM, we present Distance-Aware IBM (DA-IBM) where determines the users’ weight based on the geographical distance from the Rumor Targeted Location (RTL). In order to handle non-identical weight for users, we develop a sampling-based method called NU-IBM-Solver. Through attentively analyzing the sample size, our proposed method is able to return a (1−1/e−ϵ) approximation solution similar to greedy algorithm. As well as theoretical analysis, we perform extensive experiments over four real-world networks in various conditions. The evaluations also confirm that NU-IBM-Solver is similar to greedy in terms of effectiveness, while it is thousands of times faster. •For the first time, Non-Uniform Influence Blocking Maximization problem is introduced.•A sampling-based approach with theoretical guarantee and practical runtime efficiency is proposed.•Tested with four real-world networks with extensive experiments.•The quality of the solution has been shown both experimentally and theoretically.