Composite Behavioural Modelling for Identity Theft Detection in Online Social Networks

Abstract

The rapid growth of online social networks has significantly transformed the way people communicate, share information, and interact across the world. Popular social networking platforms such as Facebook, Instagram, X (Twitter), LinkedIn, and other online communities store vast amounts of personal information, making them attractive targets for cybercriminals. One of the most serious security threats in these platforms is identity theft, where attackers gain unauthorized access to legitimate user accounts or create fake profiles by stealing personal information and impersonating genuine users. Such attacks can lead to financial fraud, privacy violations, misinformation, reputation damage, and unauthorized access to confidential information. Traditional identity theft detection techniques mainly rely on password authentication, rule-based security mechanisms, or manual account verification, which are often insufficient to detect sophisticated attacks and evolving behavioral patterns of cybercriminals. This project proposes a Composite Behavioral Modeling for Identity Theft Detection in Online Social Networks, an intelligent security framework that utilizes Artificial Intelligence (AI), Machine Learning (ML), and behavioral analytics to identify compromised or fake user accounts by continuously monitoring multiple behavioral characteristics. The proposed system collects and analyzes various user activities, including login patterns, device information, geographical locations, browsing behavior, posting frequency, messaging habits, friend interactions, profile modifications, and session activities to construct a comprehensive behavioral profile for each user. Instead of depending on a single security feature, the system combines multiple behavioral attributes through composite behavioral modeling to accurately distinguish between legitimate and suspicious activities. Machine learning algorithms such as Random Forest, Support Vector Machine (SVM), XGBoost, Decision Tree, Logistic Regression, and anomaly detection techniques are employed to classify user behavior and identify identity theft attempts with high accuracy. The proposed system aims to improve cybersecurity by providing early detection of suspicious account activities, reducing false alarms, enhancing user privacy, minimizing financial losses, and strengthening trust in online social networking platforms through intelligent behavioral analysis.