Proactive rumor control in online networks

Abstract

The spread of rumors through online networks not only threatens the public safety but also results in loss of financial property. To serve as a reliable platform for spreading critical information, many work study the problem of rumor control which aims at limiting the pernicious influence of rumors. These methods, however, only assume that users are passive receivers of rumors even if the users can browse the rumors on their own. To overcome this issue, in this paper we study the rumor spread from a proactive perspective and introduce a novel rumor control problem, called users’ B rowsing based rU mor blocK (BUK). Given a rumor set R, BUK can be summarized as targeting k nodes as ‘protectors’ to save nodes in G from being influenced by R as many as possible. Different with the previous studies, BUK considers that the rumors spread via users’ browsing behaviors, and models the propagation based on the random walk model. Theoretical analysis shows that the problem of BUK is submodular, and we propose two greedy algorithms that can approximate BUK within a ratio of (1 − 1/e). However, both of them consume high spaces and thereby cannot be applied to very large networks. Therefore, we further propose a ranking based method RanSel to solve BUK heuristically, which only consumes a linear space to the graph size. The experiments reveal that the effectiveness of our methods outperforms the baseline by 6% to 59.2%, and our methods can achieve such an effective result in reasonable time.