SecureGait: A 3D-CNN Framework for Enhanced Biometric Authentication via Gait Recognition

Secure access to sensitive assets, such as bank lockers, demands highly reliable biometric authentication. Traditional methods like fingerprint or facial recognition, while effective, have limitations in non-intrusive scenarios. Gait recognition, based on identifying unique walking patterns, offers a promising alternative for secure, contactless authentication. Prior research has explored gait recognition using various methods, including Gait Energy Image (GEI) and deep learning-based models, yet challenges remain in adapting to varied environments, clothing, and angles. In this study, we present SecureGait, a 3D-CNN model designed to enhance gait recognition accuracy by capturing both spatial and temporal features. Trained on the CASIA-B dataset, the model achieved high accuracy, balanced error rates, and rapid inference times, demonstrating robustness across diverse conditions. These results suggest SecureGait’s potential for high-security applications, advancing gait recognition’s role in authentication solutions. Future work aims to expand the model’s adaptability and deploy ability across real-world applications.