Lightweight gait based authentication technique for IoT using subconscious level activities

With the rapid growth of IoT market, it is expected that a large number of IoT devices will be deployed in the future networked systems. However, traditional user authentication techniques may be too heavy and not be applicable to the IoT devices that have low computation and communication resources. To address such potential limitation, we propose a light-weight user authentication technique for IoT systems, called Li-GAT (Lightweight Gait Authentication Technique) that exploits various information collected from IoT devices, namely the subconscious level of user activities, to effectively authenticate users with high accuracy while reducing the resource consumption. In Li-GAT, we authenticate users by extracting and identifying different walking patterns of users (gait). We implement our technique on an Android platform to collect and analyze the accelerometer data from different users. The user authentication is done on a selected number of features using various machine learning classifiers. Our experiment results show that Li-GAT successfully authenticates users with high accuracy (96.69%) comparable to the existing techniques while using only the half number of features.