Internet of Things (IoT) Data Analytics for User Authentication & Activity Recognition

Abstract

Over the past few years, rapid progress in the technologies such as the Internet of Things (IoT), wireless sensor networks (WSNs), wearable devices, and remote mobile applications have provided personal data assets. IoT devices generate large amounts of data that provide an opening for assisting people in many ways with diverse range of potential applications. Security and privacy of users and devices is the central requirement in IoT-based applications. The uses of sensors data for human activity recognition and biometrics are increasing tremendously due to their potential applications in smart homes, smart cities, and healthcare applications. The traditional authentication mechanisms such as login-password are no longer supportable for IoT systems. Existing biometric methods are incorporated with additional sensors-based identification techniques of remote user recognition to improve the performance. IoT sensor data-based authentication methods provide an extra layer of security and user privacy. Due to the instability of IoT-sensor data, the authentication techniques need to extract a large number of features to produce highly-accurate results. The limited capabilities of computation, communication, storage, and the small battery power of IoT devices further make its implementation hard. The IoT device's data contain sensitive information, ensuring protection and confidentiality is essential. The IoT applications require continuous monitoring and control. The current authentication methods do not offer security after the point of entry. Sensors-based HAR has provided a basis for such resolution. Therefore, reliable authentication using human activities-based sensor data is an open research challenge. The information of human activities context assists in improving the performance of user authentication. We have introduced an IoT sensor data-based analytics framework utilizing the sensor data of a person’s smart devices to build unique user authentication models. These models include biometric measurements such as acceleration movement, heart rate, glucose level, and oxygen saturation level along with motion sensors, i.e., accelerometer. We validate the proposed approach on publically available real data sets. We also introduce a threat model for the proposed authentication scheme. There is no existing work that performed the users and activities recognition simultaneously. We contributed using the context-aware activity and user recognition method. The experimental analysis has achieved an accuracy of 99% for user authentication and 95% for activity recognition with a minimum set of data and features. The method is suitable for resource-constrained IoT devices with low energy consumption and less computational cost.