Trust management for IoT devices based on federated learning and blockchain

The rapid growth of IoT devices and the increasing demand for device interaction between different network partitions have significantly pressured IoT device management. To realize cross-domain trust integration between different partitioned devices, trust management becomes the key technology to realize cross-domain communication. However, trust management heavily relies on third-party entities, posing centralized risks. Therefore, we propose an IoT device trust management system based on federated learning and blockchain. By utilizing federated learning to assess device reputation ratings while safeguarding their privacy, the system stores reputation assessment results on the blockchain for authenticity and accuracy. The system's decentralization is achieved using blockchain instead of a central server in federated learning. Additionally, we introduce a weighted aggregation model based on device attributes to obtain a more precise global model through weighted aggregation of local models in federated learning. Experimental results using a simulated dataset reflecting device characteristics demonstrate a device evaluation accuracy of 90.2%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$90.2\%$$\end{document}, validating the system's effectiveness and feasibility.