Trust Hole Identification in Signed Networks

Abstract

In the trust-centric context of signed networks, the social links among users are associated with specific polarities to denote the attitudes (trust vs distrust) among the users. Different from traditional unsigned social networks, the diffusion of information in signed networks can be affected by the link polarities and users’ positions significantly. In this paper, a new concept called “ trust hole ” is introduced to characterize the advantages of specific users’ positions in signed networks. To uncover the trust holes, a novel trust hole detection framework named “ S ocial C ommunity based t R ust h OL e exp L oration” ( Scroll ) is proposed in this paper. Framework Scroll is based on the signed community detection technique. By removing the potential trust hole candidates, Scroll aims at maximizing the community detection cost drop to identify the optimal set of trust holes. Extensive experiments have been done on real-world signed network datasets to show the effectiveness of Scroll .