A lightweight machine learning-based authentication framework for smart IoT devices

Abstract

The Internet of Things (IoT) is the next generation plethora of interconnected devices that includes sensors, actuators, etc. and that can provide personalized services such as healthcare, security, and surveillance. The quality of our daily lives is improved by the IoT through pervasive computation and communication. Innumerable devices are being connected each day to IoT applications. Although the quality of our lives is enhanced by the IoT, IoT applications also cause serious challenges in securing networks and data in transit. Existing security solutions, such as password-based two-factor authentication and traditional biometric template-based authentication, can be challenged because of several threats that affect the reliability and efficiency of the entire system. Hence, there is a need for a highly secure authentication mechanism such as the Cancelable Biometric System (CBS). In essence, the CBS is a biometric template protection scheme that operates based on repeated distortions/transformations at the feature/signal level. Therefore, in this paper, we propose a framework for a cloud-based lightweight cancelable biometric authentication system. Findings from our study are used to demonstrate the potential for the proposed approach to be deployed in real-world settings (i.e., the capability to authenticate client devices with high accuracy and minimal overhead without affecting the security of the sensitive biometric templates in the cloud environment). Both theoretical and experimental analyses suggest that the proposed approach has a minimal equal error rate compared with those of the state-of-the-art techniques. Moreover, the proposed approach has been proven to consume less time, making it suitable for IoT environments.