INTEGRATING SWIN TRANSFORMER AND AES-256 ENCRYPTION FOR A HIGH-SECURITY BIOMETRIC IDENTIFICATION
DOI:
https://doi.org/10.61591/jslhu.26.1097Keywords:
Facial Expression Recognition;, Principal Component Analysis;, Local Binary Patterns;, Support Vector Machine;, Euclidean Distance.Abstract
This article addresses the core challenges facing modern biometric identification systems: the limited global context capture of traditional Convolutional Neural Networks (CNNs), the rise of sophisticated spoofing attacks, and the risk of sensitive data breaches. We propose a high-security system architecture that integrates Swin Transformer for facial feature extraction, YOLOv8 for high-speed object detection, and MiniFASNet for anti-spoofing based on Fourier frequency domain analysis. A primary focus of this research is data privacy, achieved through the implementation of an AES-256 encryption mechanism for biometric feature vectors. The model was trained using a transfer learning strategy with a frozen backbone on the VGGFace2 dataset and validated against real-world data. Experimental results demonstrate that the system achieves an AUC of 0.794, an anti-spoofing accuracy of 96.5%, and a real-time processing speed of approximately 36ms per face on consumer-grade hardware. This study confirms the feasibility of combining advanced Vision Transformer architectures with robust security protocols under limited computational resources.
References
M. Elgendy, Deep Learning for Vision Systems. Manning Publications, 2020.
J. Brownlee, Deep Learning for Computer Vision: Image Classification, Object Detection and Face Recognition in Python. Machine Learning Mastery, 2019.
Z. Yu et al., "Rethinking Vision Transformer and Masked Autoencoder in Multimodal Face Anti-Spoofing," International Journal of Computer Vision, vol. 132, pp. 5217–5238, 2024.
M. Rodrigo, C. Cuevas, and N. García, "Comprehensive comparison between vision transformers and convolutional neural networks for face recognition tasks," Scientific Reports, vol. 14, art. no. 21392, 2024.
R. Gao and T. Bourlai, "On Designing a SwinIris Transformer Based Iris Recognition System," IEEE Access, vol. 12, pp. 30737–30752, 2024.
S. Sharhrah et al., "Multimodal Biometric Authentication using Convolutional Neural Network, Swin Transformer, and Multi-Head Attention with Global Max Pooling," ITM Web of Conferences, vol. 79, art. no. 01052, 2025.
R. Garg, P. Pathak, and M. P. Singh, "A multimodal biometric recognition system based on Fingerprints, Iris and ECG via Swin Transformer and CNN Model," Systems and Soft Computing, vol. 7, art. no. 200369, 2025.
S. Nagaraju et al., "Biometric key generation and multi round AES crypto system for improved security," Measurement: Sensors, vol. 30, art. no. 100931, 2023.
