An Advanced Machine Learning Model for Detecting Synthetic Identity Fraud in E-Commerce Platforms

Abstract

The rapid growth of e-commerce has created unprecedented opportunities for global trade, yet it has simultaneously exposed platforms to sophisticated fraudulent activities, among which synthetic identity fraud (SIF) represents a particularly insidious threat. SIF involves the creation of fictitious identities, combining real and fabricated information, to exploit digital commerce systems, often evading traditional identity verification mechanisms. The detection of such fraudulent behavior presents unique challenges due to the hybrid nature of synthetic identities, the high-dimensionality of user data, and the dynamic evolution of fraudulent tactics. This study presents an advanced machine learning (ML) model designed to identify synthetic identities in e-commerce platforms by leveraging multi-layered feature extraction, ensemble learning strategies, and anomaly detection techniques. The proposed model integrates supervised and unsupervised learning approaches to enhance detection accuracy while reducing false positives. In addition, it incorporates adaptive learning mechanisms that adjust to evolving fraud patterns, thereby increasing robustness against adversarial attempts. Empirical evaluation on large-scale e-commerce datasets demonstrates the model’s effectiveness in distinguishing synthetic from legitimate users, offering a scalable solution for real-world application. The findings contribute to both the academic understanding of synthetic identity fraud and practical solutions for enhancing e-commerce platform security.