"Hybrid Model for Facial Life and Impersonation Detection Using Multimodal Deep Learning"

Abstract

"Currently, spoofing is a problem that affects many people, becoming one of the main techniques used by cybercriminals and the lack of mechanisms to identify spoofing frauds causes many people to be victims of various digital scams. Therefore, the objective of this study is to identify face life detection of images to combat this problem by combining deep learning techniques and graphene neural networks. The methodology used was based on 5 phases as obtained data set, preprocessing, training with deep learning models (VIT, GCN), model evaluation (accuracy, precision, recall, F1-Score, AUC-ROC), classification result. The best result was obtained using a hybrid model (ConVit + GCN) with 91.84% accuracy, 94.17% accuracy, 89.21% recall, 91.62% F1-Score, 98.12% AUC-ROC. The results show a model that combines 2 techniques based on neural networks capable of identifying and combating spoofing. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026."