INTEGRATING BLOCKCHAIN AND MACHINE LEARNING FOR ROBUST IOT DATA INTEGRITY IN CRITICAL INFRASTRUCTURES

Abstract

The emerging utilization of Internet of Things (IoT) devices in some critical infrastructures like smart grids and healthcare systems poses urgent issues of data integrity, security, and trust. Traditional security approaches break down in the face of major, sophisticated attacks like data injection, botnet, and DDoS attacks. This paper introduces a hybrid system which combines blockchain and machine learning to improve the integrity of IoT data and detect unexpected anomalies. Blockchain guarantees immutability and trust by decentralization and smart contracts and machine learning models are used to specialize in anomaly or malicious behavior of IoT telemetry. The proposed approach is evaluated using the TON_IoT dataset [6] (2020) in which the real-world IoT telemetry data together with the corresponding attack scenarios are offered. Experimental results show that the proposed method is able to achieve ultra-detector performance with a much higher detection accuracy and a much lower number of false positives than existing independent methods. This work showcases the possibility of leveraging blockchain-inspired technologies with knowledge-driven learning to establish robust, secure, scalable data streams for CIs.