SentinelNet: Adaptive Blockchain Security for Wireless Sensor Networks

Abstract

Wireless Sensor Networks (WSNs) are employed in numerous applications related to the Internet of Things (IoT) such as smart healthcare, industrial monitoring, intelligent infrastructure, etc., but their resource constraints and use of wireless communications makes them susceptible to attacks like Sybil attacks, replay attacks, eavesdropping, and data manipulation. This paper introduces SentinelNet, which is a blockchain-based security system with an adaptive design that integrates AES-256-CBC encryption technology, SHA-512 chain-linked verification of integrity, machine learning-based attack detection, and logging on Ethereum smart contracts in one framework. The attack detection system utilized in this framework employs the Random Forest classification system, the Isolation forest system for detecting anomalies, and the dynamic evaluation of trust score in order to identify the attacks without using too much computing power. Tests were conducted using 50 sensor nodes and 1,000 sensing cycles in the experiments. The experiments conducted produced F1 scores of 97.8 and 98.1% for the Sybil attack and the replay attack respectively with an additional energy expenditure of 3.2% on the LEACH model.