Suppressing the Spread of Fake News Over the Social Web of Things: An Influence Maximization-Based Supervised Approach

Abstract

With the continuous rise of various social media, the security problems caused by disseminating news in social networks have become increasingly prominent. Among them, the dissemination of false news has brought a significant threat to the security of cyberspace. A method based on influence maximization is proposed to change the premise of the false news in network space. This article employed the Louvain clustered local degree centrality (LCLD) and random maximum degree (RMD) algorithm to extract the most influential node set via influence maximization. Subsequently, TextCNN (convolutional neural network) was used to classify and identify false news and filter out several critical points in the node set. Then, the modified propagation network was repredicted by the prediction model. As a result, the spread of false news was significantly suppressed compared with the network before modification. Finally, the verification was carried out on Buzz Feed News’s real dataset. First, the information-based cascaded prediction model can more accurately fit the actual spread. Then the modified network was input into the prediction model for prediction, and the results show that the spread of false news can be suppressed. Finally, the influence maximization algorithm deletes several nodes containing incorrect information. It can effectively suppress the spread of false news, thus verifying the effectiveness of the proposed method.