Securing eCommerce Payments with DeepFake-Aware Facial Biometrics

Abstract

The rapid digitization of financial services has increased the reliance on biometric systems, particularly facial recognition, for seamless identity verification in eCommerce transactions. However, the emergence of sophisticated DeepFake techniques has introduced a substantial new attack surface for adversaries to manipulate facial biometrics and bypass authentication systems. This paper investigates a DeepFake-aware biometric verification framework designed to harden payment systems against synthetic facial forgeries. The study explores the vulnerabilities of existing face-ID mechanisms, evaluates modern DeepFake generation pipelines, and proposes a hybrid detection architecture leveraging Convolutional Neural Networks (CNNs) combined with Vision Transformers (ViTs). Experimental evaluations are conducted using publicly available DeepFake datasets and custom eCommerce-oriented attack samples. The results demonstrate that traditional facial recognition systems experience drastic drops in accuracy under adversarial synthetic faces, whereas the proposed DeepFake-aware facial biometric framework significantly enhances fraudulent detection rates while maintaining high verification reliability. This paper contributes actionable insights for secure digital payments, offering a scalable and robust defense mechanism for real-world financial platforms.