Secure facial biometric authentication in smart cities using multimodal methodology

Aanjankumar Sureshkumar ¹, Malathy Sathyamoorthy ², Rajesh Kumar Dhanaraj ³

⁰School of Computing Science and Engineering, VIT Bhopal University, Bhopal-Indore Highway, Kothrikalan, Sehore, 466114, Madhya Pradesh, India.

Scientific Reports +

|

December 29, 2025

View abstract on PubMed

Summary

This study introduces a multimodal deep learning model for secure facial biometric authentication in smart cities. The system combines Convolutional Neural Network (CNN) and ResNet-50 with ElGamal cryptography, achieving 97.1% accuracy against spoofing and enhancing data security.

Area Of Science

Computer Science
Artificial Intelligence
Cybersecurity

Background

Facial biometric security is critical in smart cities for protecting citizen data and preventing unauthorized access.
Existing systems face challenges with spoofing attacks and secure data transmission.
Need for robust authentication methods that combine feature extraction and cryptographic security.

Purpose Of The Study

To propose a multimodal deep learning model integrated with a cryptographic framework for enhanced facial biometric authentication.
To secure facial data against spoofing attacks and ensure privacy during transmission in smart city networks.
To evaluate the model's performance and compare it with traditional methods.

Main Methods

Utilized a multimodal deep learning approach combining Convolutional Neural Network (CNN) for low-level feature extraction and Residual Network (ResNet-50) for high-level semantic pattern identification.
Integrated ElGamal cryptography to secure extracted facial features and ensure data privacy.
Trained and evaluated the model on the CelebA Faces dataset.

Main Results

Achieved a facial mapping prediction accuracy of 97.1% with a low mean score loss of 0.04.
Demonstrated superior performance compared to traditional models: 1.2% higher accuracy than CNN, 2.2% higher than ResNet-50, and 1.1% higher than the Brakerski-Gentry-Vaikuntanathan algorithm.
Effectively handled spoofing attacks and ensured secure data transmission.

Conclusions

The proposed fused multimodal approach significantly enhances facial biometric security in smart city environments.
The combination of CNN, ResNet-50, and ElGamal cryptography provides a robust solution for preventing unauthorized access and ensuring data privacy.
The model is well-suited for futuristic smart city applications requiring advanced security features.

Keywords:

Convolutional neural network Deep learning ElGamal Face authentication ResNet-50 Spoofing attacks