Blockchain-Based Federated Learning Using Machine Learning Classifier for IoT Intrusion Detection System

As the number of users and applications in IoT networks increases, concerns about cyber attacks are growing. IoT systems and other industries have significantly increased the adoption of machine learning-based intrusion detection systems (IDS) to defend against cyber attacks. Due to computing costs and privacy issues, using traditional centralized machine learning (ML) approaches to analyze data stored on a single central server is not a reasonable solution. The most promising strategy to address this issue is federated learning (FL), which updates learning on edge devices without exposing personal data to a central server. Combining federated learning with blockchain (BC) can provide greater benefits in solving this problem due to enhanced levels of privacy and security. In this paper, we propose a blockchain-based architecture (BFLSID) to support federated learning and improve intrusion detection in IoT networks. We have also developed a distributed storage system that integrates InterPlanetary File System (IPFS) and Hyperledger Fabric to ensure the security of model training and detection results.