Privacy-Preserving Blockchain-Based Federated Learning for IoT Devices

Home appliance manufacturers strive to obtain feedback from users improve their products and services build a smart home system. To help develop system, we design federated learning (FL) system leveraging reputation mechanism assist train machine model based on customers’ data. Then, can predict requirements consumption behaviors in the future. The working flow of includes two stages: first stage, customers initial provided by manufacturer using both mobile phone mobile-edge computing (MEC) server. Customers collect data various appliances phones, then they download with local After deriving models, sign models send them blockchain. In case or are malicious, use blockchain replace centralized aggregator traditional FL Since records untampered, malicious manufacturers’ activities traceable. second select organizations as miners for calculating averaged received customers. By end crowdsourcing task, one miners, who is selected temporary leader, uploads protect privacy test accuracy, enforce differential (DP) extracted features propose new normalization technique. We experimentally demonstrate that our technique outperforms batch when under DP protection. addition, attract more participate an incentive award participants.