Enhancing security in online social networks: introducing the DeepSybil model for Sybil attack detection

Online Social Networks (OSNs) have become increasingly popular platforms for personal, professional, and social networking. However, the rise of fraudulent events such as fake news and rumors has created security threats within OSNs, leading to significant impacts. Among the major threats are Sybil attacks, where fake accounts, known as Sybils, are generated within OSNs to carry out various malicious activities. To address this issue, this paper proposes the DeepSybil model, which utilizes the Gannet Optimization Algorithm (GOA) and Graph Attention Networks (GATs) to detect Sybil attacks in OSNs. The proposed model starts by gathering various input information using the Social Network Fake Account (SNFA) dataset. This data is then preprocessed to eliminate redundant and irrelevant data points. Next, meaningful features are extracted to capture user behaviors and interactions, while an adjacency matrix represents the links between users to analyze the network structure of OSNs. Categorical features are encoded into binary features using One-hot encoding. The dataset is divided into training and testing sets to evaluate the model's efficiency. The GOA algorithm is employed to optimize the parameters of the proposed model, thereby improving the accuracy of Sybil attack detection. GATs leverage the attention mechanism to accurately identify Sybil attacks within OSNs. The model outputs a probability indicating the likelihood of an account being a Sybil. To evaluate the effectiveness of the proposed model, various performance metrics, including accuracy, precision, F1-score, recall, Receiver Operating Characteristic curve (ROC), and specificity, are employed. The results demonstrate that the proposed model achieves a high accuracy of 96.8%, outperforming existing methods. Thus, the proposed DeepSybil model proves to be a highly suitable solution for accurately detecting Sybil attacks in OSNs.