Lightweight DeepFake-Resistant Facial Recognition for Mobile eCommerce

Abstract

The rapid adoption of mobile eCommerce platforms has amplified the need for secure yet lightweight authentication mechanisms. Traditional facial recognition systems often fail under resource constraints of mobile devices and are increasingly vulnerable to DeepFake attacks, which can manipulate facial features to bypass security. This paper introduces a lightweight DeepFake-resistant facial recognition framework specifically designed for mobile eCommerce applications. The proposed system combines efficient convolutional neural network architectures with real-time liveness detection and feature-level DeepFake detection techniques to achieve robust authentication without imposing high computational costs. Experimental evaluations on publicly available DeepFake datasets and mobile device simulations demonstrate high accuracy, minimal latency, and strong resistance to adversarial attacks. This work offers a practical, scalable solution for mobile security in digital transactions.