Multi-source localization on complex networks based on community detection

School of Cyber Engineering, Xidian University - Xi'an, Shaanxi 710071, China

^(a) E-mail: yuanshanxiaowu@qq.com
^(b) E-mail: wangchao@xidian.edu.cn (corresponding author)

Received: 22 March 2022
Accepted: 28 February 2023

Abstract

Source localization is a crucial task on complex networks to understand the transmission mechanism and eliminate the malignant influence of diffusion processes. The previous works mainly detect multiple sources from a global perspective within a network. In this paper, we introduce community detection to source localization, aiming to address source localization from a local view. Specifically, a community detection-based multi-source localization method CDSL is proposed. Experiments on both several real-world and synthetic networks demonstrate that CDSL has higher accuracy and needs less runtime than the existing source localization methods. We also explore how different observation selection strategies affect the performance of CDSL and find out that selecting the nodes with a small degree as observers performs best.