Modeling cascading failures with the crisis of trust in social networks
Chengqi Yi,Yuanyuan Bao,Jingchi Jiang,Yibo Xue
DOI: https://doi.org/10.1016/j.physa.2015.05.030
2015-01-01
Abstract:In social networks, some friends often post or disseminate malicious information, such as advertising messages, informal overseas purchasing messages, illegal messages, or rumors. Too much malicious information may cause a feeling of intense annoyance. When the feeling exceeds a certain threshold, it will lead social network users to distrust these friends, which we call the crisis of trust. The crisis of trust in social networks has already become a universal concern and an urgent unsolved problem. As a result of the crisis of trust, users will cut off their relationships with some of their untrustworthy friends. Once a few of these relationships are made unavailable, it is likely that other friends will decline trust, and a large portion of the social network will be influenced. The phenomenon in which the unavailability of a few relationships will trigger the failure of successive relationships is known as cascading failure dynamics. To our best knowledge, no one has formally proposed cascading failures dynamics with the crisis of trust in social networks. In this paper, we address this potential issue, quantify the trust between two users based on user similarity, and model the minimum tolerance with a nonlinear equation. Furthermore, we construct the processes of cascading failures dynamics by considering the unique features of social networks. Based on real social network datasets (Sina Weibo, Facebook and Twitter), we adopt two attack strategies (the highest trust attack (HT) and the lowest trust attack (LT)) to evaluate the proposed dynamics and to further analyze the changes of the topology, connectivity, cascading time and cascade effect under the above attacks. We numerically find that the sparse and inhomogeneous network structure in our cascading model can better improve the robustness of social networks than the dense and homogeneous structure. However, the network structure that seems like ripples is more vulnerable than the other two network structures. Our findings will be useful in further guiding the construction of social networks to effectively avoid the cascading propagation with the crisis of trust. Some research results can help social network service providers to avoid severe cascading failures.