A Biometric-Driven E-Commerce Authentication Model: Integrating Client-Side Liveness Detection with MERN Architecture for Fraud Mitigation
The pervasive vulnerability of password-based authentication to credential stuffing, phishing, and identity-spoofing attacks represents a persistent and unresolved threat to electronic commerce security. This paper presents the design, implementation, and empirical evaluation of a biometric-driven e-commerce authentication model that integrates client-side real-time liveness detection with a full-stack MERN architecture comprising MongoDB, Express.js, React.js, and Node.js. The system employs Face-api.js for in-browser facial landmark detection, 128-dimensional descriptor extraction, and liveness verification, transmitting only encrypted numerical embeddings to the backend rather than raw biometric imagery. Authentication logic is routed through RESTful middleware-validated endpoints, including a dedicated /api/facial-login route, with session management enforced via JSON Web Tokens. Empirical evaluation across five user cohorts (U1–U5) demonstrated a peak recognition accuracy of 97.8%, a False Acceptance Rate (FAR) of 0.02, and a False Rejection Rate (FRR) of 0.02 at the optimal threshold configuration. All authentication cycles completed within five seconds, with a minimum latency of 2.91 seconds. Cyber risk assessment confirmed a 50% reduction in unauthorised access attempts and a 16.67% reduction in fraud incidents post-deployment. The architecture demonstrates that client-side biometric processing, when properly secured through descriptor encryption and middleware validation, can deliver high-performance authentication without centralized biometric data storage, substantially mitigating both security and privacy risks in online commerce environments.
Download Paper