An Ensemble-Oriented Hybrid XNS Approach for Detecting Abnormal Behavior in Social Networks

Resumen

The rapid rise of online social networks (OSNs) has significantly reshaped how people communicate and share information, but it has also brought growing concerns around user security, privacy, and overall well-being. Identifying abnormal behavior—such as the spread of fake news, sharing malicious links, or engaging in harmful online activities—has become increasingly important for ensuring safer digital environments. To address these challenges, this study introduces an ensemble-oriented hybrid XNS framework that combines XGBoost, Support Vector Machine (SVM), and Naïve Bayesian Classifier (NBC) in a cascaded architecture, designed to improve detection accuracy and robustness. The proposed framework draws on a rich set of features extracted from both user profiles and user-generated content, allowing for a more comprehensive understanding of user behavior patterns. It is trained and tested using a combination of synthetic and real-world datasets, ensuring that the model remains adaptable and reliable across diverse scenarios. In addition to conventional anomaly detection, the framework also places special emphasis on identifying users who may show signs of depression or suicidality, thereby supporting early intervention and promoting mental health awareness. Experimental results demonstrate that the model achieves an accuracy of approximately 97%, outperforming traditional methods and highlighting its potential as a scalable and efficient solution for enhancing both cybersecurity and user well-being in modern social networks.