Evaluating Data Trust in Blockchain-Based IoT Systems Using Machine Learning Techniques

The convergence of blockchain and Internet of Things (IoT) has become increasingly prevalent recently, as it addresses challenges such as single point of failure and security concerns associated with IoT. Blockchain offers immutable data storage, availability, and transparency, but a significant drawback lies in its inability to verify the truthfulness of the data stored on it. State-of-the-art systems attempting to mitigate this concern often rely on conventional reputation-based approaches, which predominantly evaluate historical data from sensors and neglect the critical assessment of data in real-time. Furthermore, there is limited research on incorporating machine learning (ML) methods to enhance data trustworthiness in blockchain systems. This paper proposes a novel ML-based trust assessment approach that takes into account both historical reputation and real-time data trust-worthiness. Our approach integrates multiple ML models within a blockchain framework using edge servers and validators, effectively functioning as a distributed ensemble to enhance classification accuracy, and contributing to more accurate reputation score calculations. Our results demonstrate significant accuracy gains in distinguishing trustworthy and untrustworthy IoT sensor data in blockchain networks.