Random Full-Order-Coverage Based Rapid Source Localization With Limited Observations for Large-Scale Networks

Dongpeng Hou, Chao Gao, Zhen Wang, Xiaoyu Li, Xuelong Li

Abstract: The rapid spread of misinformation in social media presents significant threats to society, highlighting the importance of early inference of the diffusion source to minimize potential losses. Although sensor-based methods have proven effective in source localization, their reliance on sufficient information from all sensors restricts their ability to accurately identify the source with limited data from a few sensors, thereby limiting their application in early propagation scenarios. To address these challenges, this paper introduces a novel method called r andom f ull-order-coverage based r apid s ource l ocalization (RF-RSL). RF-RSL improves the g reedy-based s trategy (GS) in a random deployment way to quickly provide extensive coverage of deployed sensors over a wide area, followed by the l imited-information-oriented s trategy (LS) for source inference with an early response mechanism. Specifically, LS incorporates a quick preprocessing step to eliminate invalid candidates and a novel source estimator for precise source identification. The experiments demonstrate that RF-RSL consistently outperforms the best baseline by at least 5% and exhibits exceptional advantages of up to 30% when deployed with fewer sensors. Moreover, RF-RSL showcases a remarkable speed advantage of over 10 times compared to the best baseline in large-scale networks.