Detection of Multi-Account Abuse in E-Commerce Systems Using Behavioral Biometrics

Abstract

The proliferation of e-commerce platforms has increased both convenience for consumers and opportunities for malicious activities, particularly multi-account abuse, where a single user manipulates multiple accounts to exploit promotions, bypass restrictions, or commit fraud. Traditional detection mechanisms, such as IP tracking and device fingerprinting, are increasingly ineffective against sophisticated attacks that employ automation, VPNs, or virtual devices. This research proposes a behavioral biometric-based framework to detect multi-account abuse by analyzing user interaction patterns, including typing dynamics, mouse movements, click frequency, and navigation sequences. The system extracts statistical and temporal features from user sessions, normalizes and aggregates them into behavioral vectors, and applies machine learning algorithms—Random Forests, Support Vector Machines, and Neural Networks—for classification. A behavioral similarity analysis identifies accounts with highly correlated interaction patterns, indicating potential coordinated abuse. Experimental results on a synthetic e-commerce behavioral dataset demonstrate that the proposed approach achieves high detection accuracy (>95%), strong precision, recall, and ROC-AUC scores, outperforming traditional rule-based methods. This study highlights the efficacy of behavioral biometrics in enhancing fraud detection systems while preserving a seamless user experience, providing a foundation for real-time and scalable anti-fraud solutions in e-commerce.