Graph Neural Network Based Approach for Restraining Misinformation Propagation in Online Social Networks

This paper addresses the problem of misinformation spread within online social networks (OSNs), a critical concern impacting public opinion and societal stability. We introduce a sophisticated Graph Neural Network-based framework designed to effectively curb misinformation propagation in OSNs. This framework is composed of three integral components: the Training Dataset Generator, the Graph Convolutional Network (GCN) Model Trainer, and the Influential Node Identifier. The Training Dataset Generator is crucial for calculating the acceptance rate of misinformation across nodes, thereby setting the stage for targeted interventions. The GCN Model Trainer utilizes this data to predict acceptance rates of misinformation for each node, identifying the most influential ones likely to spread misinformation. Finally, the Influential Node Identifier leverages these predictions to apply a strategic blocking method on key nodes. Employing a trained GCN model, we introduce a greedy algorithm aimed at selecting the most susceptible nodes for effective misinformation control. Our approach, when compared against baseline methodologies, demonstrates superior performance in mitigating the spread of misinformation.