Identifying Fake Accounts on Social Media Using Graph Neural Networks

We proposed a multimodal machine learning approach to identify fake accounts on social media platforms. First, we employed Support Vector Machine (SVM) to analyze the activity status of individual accounts and detect potentially anomalous account behaviors. In addition, we trained Graph Neural Networks (GNNs) to explore the relationship structure among accounts. Through Graph Neural Networks, we captured the complex interaction patterns and associations among social media accounts. Finally, by integrating both models, the accuracy of identifying fake accounts was increased. Through such a machine learning approach, the relationships among social media accounts were analyzed effectively. The proposed model achieved an accuracy rate of 81.1 % for fake account detection, outperforming that of a pure SVM model.