Machine Learning Applications in Behavioural Authentication Data Using Gait Analysis

Abstract

In an era marked by the increasing demand for secure and user-friendly authentication methods, the convergence of behavioral biometrics and machine learning offers a compelling solution. The proposed methodology elucidates the process of data collection, emphasizing the significance of diverse and contextually varied gait datasets. Feature extraction techniques, encompassing spatiotemporal parameters and joint kinematics, are detailed to facilitate the translation of raw gait data into informative features. Machine learning model selection, a pivotal aspect of the methodology, includes a strategic choice of algorithms tailored to gait analysis, with considerations spanning model complexity, interpretability, and performance. The authors present a review of pertinent literature, a robust methodology for data collection and analysis, experimental results, and a discussion of findings. This study demonstrates the potential of machine learning in leveraging gait analysis as a reliable and unobtrusive means of user identification with applications ranging from access control to healthcare monitoring.