Enhancing Smartphone-Based Gait Authentication: A Comprehensive Review of Methods and the Potential of Zero-Shot Learning

Gait authentication, using smartphone sensors like accelerometers and gyroscopes, provides a non-intrusive and user-friendly alternative to conventional methods like passwords and PINs. However, despite its potential, gait-based authentication faces significant challenges, including gait variability due to walking conditions, sensor placement sensitivity, susceptibility to spoofing attacks, and the need for large labeled datasets for training deep learning models. This study first critically reviews existing smartphone-based gait authentication techniques, quoting their strengths, limitations, and challenges. The paper then explores the integration of Zero-Shot Learning (ZSL) to enhance accuracy and scalability and discusses future research directions for improving real-time authentication, cross-device compatibility, and privacy-aware implementations. This study aims to bridge the gap between security, adaptability, and minimal data dependency, laying the foundation for a more scalable and robust mobile authentication system.