A Machine-Learning-Based Approach for Autonomous IoT Security

Abstract

Machine learning techniques are proven valuable for the Internet of things (IoT) due to intelligent and cost-effective computing processes. In recent decades, wireless sensor network (WSN) and machine learning are integrated to give significant improvements for energy-based systems. However, resourceful routes analytic with nominal energy consumption are some demanding challenges. Moreover, WSN operates in an unpredictable space and a lot of network threats can be harmful to smart and secure data gathering. Consequently, security against such threats is another major concern for low-power sensors. Therefore, we aim to present a machine learning-based approach for autonomous IoT Security to achieve optimal energy efficiency and reliable transmissions. First, the proposed protocol optimizes network performance using a model-free Q-learning algorithm and achieves fault-tolerant data transmission. Second, it accomplishes data confidentiality against adversaries using a cryptography-based deterministic algorithm. The proposed protocol demonstrates better conclusions than other existing solutions.