IoT Data Security: An Integration of Blockchain and Federated Learning

Abstract

Technological advancement has led to a rapid increase in the growth of IoT devices leading to a vast amount of generated data. Manufacturers of such devices utilize machine learning algorithms to extract valuable insights from user data. However, this can give rise to critical issues surrounding data leakage and privacy. To tackle these issues, utilizing blockchain as a decentralized database to securely store data and employing federated learning to extract useful insights from user data can provide a viable solution. In this paper, we propose a three-layered, decentralized architecture that uses a traditional federated learning mechanism in conjunction with the Ethereum blockchain. Moreover, for data management, we use Inter-Planetary File System (IPFS) which is a peer-to-peer network used to store data in a decentralized manner. We tested our model's feasibility by using CIFAR-10 dataset and Python as the programming language with a framework for federated learning on a general purpose computer. We used Ganache_v2.5.4 and Truffle_v5.4.22 for developing smart contracts and testing and deploying them over the Ethereum blockchain.