Identify Souce of Rumor Spread in Complex Large Scale Network

Abstract

Suppose a rumor has spread in a network for some time, given a snapshot of the infected network, can we tell which node is most likely to be the rumor source? This is referred as rumor source inferring problem. Recent studies on this problem have shown appealing results if the network is tree like. However little development has been made for general large, complex network, which is what this project tries to address. Several contributions are made in this project. First of all, a rumor spreading model for general graph is proposed; second, an important property of rumor centrality (a network centrality measure tightly related to this problem) is found in regular graph with single cycle of length 3, and thus an efficient algorithm is proposed; third, extensive simulations have been conducted for various inferring techniques on both synthetic and real-world networks, by which some insights are gained; finally, a fast, stable and scalable visual simulator is implemented, which can visualize and animate the rumor spreading and source inferring process nicely.