Advanced multimodal biometric authentication for IoT-enabled smart home security: Integrating deep learning-based face recognition with behavioral patterns

This research proposed an advanced multimodal biometric authentication system for IoT-enabled smart home security. The security issues related to IoT-enabled smart home devices include, but are not limited to, unauthorized access, the risk of biometric spoofing, and reliability problems with single-modal authentication when illumination levels, user pose or position, or viewing conditions differ. The system integrates deep learning-based face recognition and gait analysis using the CASIA-B dataset. A convolutional neural network (CNN) model is employed for face recognition, while a gait energy image (GEI) is utilized for gait analysis. A novel weighted multimodal fusion approach is developed to combine results from both modalities, balancing their inputs to enhance authentication accuracy. Experiments demonstrate robust performance across various viewing angles, achieving average accuracies of 88 % for face recognition and 89 % for gait analysis. The integrated model outperforms individual modalities, achieving an overall accuracy of 92 %, highlighting the potential of multimodal systems to improve biometric security in smart homes. The implications of this study are significant for the future of residential security systems. By demonstrating the effectiveness of combining multiple biometric modalities, this research paves the way for more secure and user-friendly smart home access control systems, potentially reducing unauthorized access attempts and enhancing overall home safety in the internet of things (IoT) era.