Diffusion Containment in Complex Networks Through Collective Influence of Connections
Yang Liu,Guangbo Liang,Xi Wang,Peican Zhu,Zhen Wang
DOI: https://doi.org/10.1109/tifs.2023.3338423
IF: 7.231
2024-01-01
IEEE Transactions on Information Forensics and Security
Abstract:We study the containment of diffusion in a network immunization perspective, whose solution also plays fundamental roles in scenarios such as the inference of rumor sources and the control of malicious viral marketings. In general, the network immunization aims to suppress the giant connected component of a network by removing as fewer nodes as possible, so that the intervention of the transmission could be achieved by only a few resources. Here, rather than that and based on the fact that removing edges might be cheaper and more applicable in some scenarios, we investigate which group of edges whose removal could boost the performance of an immunization strategy more effectively. We consider both cases that the network topology is known and unknown, and thus two approaches are accordingly developed based on the Edge RelationShip (ERS) and Explosive Percolation over Partial (EPP) information. We evaluate the performance of ERS by comparing it with strategies based on the edge betweenness, the product of eigenvector centralities of the nodes connected by edges, the epidemic link equations, etc. Results on over 30 real networks show that ERS could effectively acquire far better solutions by much less computing time. We also demonstrate the performance of EPP in the circumstances of decentralized, centralized, and delayed cases. We find that the performance of EPP would be in a degree degraded by the uncertainty of inferences from individuals, inaccuracy of predictions, and delay of reactions. But in almost all cases, the developed approach can more effectively suppress a diffusion compared to the currently random strategy, especially when a tough restriction is needed or a combination with the acquaintance immunization is conducted.