Person Authentication by Gait Data from Smartphone Sensors Using Convolutional Autoencoder

Abstract

Biometric authentication is a security process that relies on the unique biological characteristics of an individual to verify who he or she is. Human gait serves as an important non invasive biometric modality for an authentication tool in various security applications. Recently due to increased use of smartphones and easy capturing of human gait characteristics by embedded smartphone sensors, human gait related activities can be utilized to develop user authentication model. In this work, a new method for user authentication from smartphone sensor data by a hybrid deep network model named convolutional autoencoder has been proposed and the performance of the model is compared with other machine learning including deep learning based techniques by simulation experiments with bench mark data sets. It is found that our proposed authetication method from smartphone sensor data with convolutional autoencoder reduces the time for authentication and also produces fair authentication accuracy and EER. It can be potentially used for person authentication in real time.