Lightweight source localization for large-scale social networks

Zhen Wang, Dongpeng Hou, Chao Gao, Xuelong Li

Abstract: The rapid diffusion of hazardous information in large-flow-based social media causes great economic losses and potential threats to society. It is crucial to infer the inner information source as early as possible to prevent further losses. However, existing localization methods wait until all deployed sensors obtain propagation information before starting source inference within a network, and hence the best opportunity to control propagation is missed. In this paper, we propose a new localization strategy based on finite deployed sensors, named Greedy-coverage-based Rapid Source Localization (GRSL), to rapidly, flexibly and accurately infer the source in the early propagation stage of large-scale networks. There are two phases in GRSL. In the first phase, the Greedy-based Strategy (GS) greedily deploys sensors to rapidly achieve wide area coverage at a low cost. In the second phase, when a propagation event within a network is observed by a part of the sensors, the Inference Strategy (IS) with an earlier response mechanism begins executing the source inference task in an earlier small infected area. Comprehensive experiments with the SOTA methods demonstrate the superior performance and robustness of GRSL in various application scenarios.