Weibo Spammer Detection Based On Social Network Digital Twin

Abstract

Users are increasingly willing to share their comments on the Internet. The popularity of Weibo has spawned spammers. Comments from spammers affect normal Internet public opinion. The traditional spammer detection methods are mainly based on the static characteristics of users and accuracies are not ideal. In this paper, we apply the parallel system framework to build a social network digital twin. The nodes of the digital twin are mapped to the nodes of the graph attention network and the relationships between nodes in the digital twin are mapped to the neighbor nodes in the graph attention network. The feature vectors of nodes are updated by the stacked graph attention layer. We take the output of the attention layer as the input of the full connection layer. The softmax classifier is used to get the classification results. In this paper, we wrote a crawler to collect the individual information and follow the relationship of 2,000 users, screened out 15 user characteristics, and manually annotated them. The experimental results show that the model we proposed has higher accuracy than the naive Bayes model and decision tree.