Design and implementation of privacy-preserving federated learning algorithm for consumer IoT

Abstract

Home appliance manufacturers are increasingly focusing on building smarter home systems by incorporating user feedback to enhance products and services. To support this, we designed a federated learning (FL) system that includes a reputation mechanism to help manufacturers leverage customer data to train machine learning models. First, it downloads the initial model provided by the manufacturer and trains it with local data. Then, it asks customers to sign their models and upload them to the blockchain. To protect customer privacy, we implemented differential privacy and introduced a new normalization technique. In addition, we also attract more customers to participate in crowdsourced FL tasks by rewarding their contributions, thereby ensuring that the datasets for model training are robust and diverse. This system not only promotes collaboration between customers and manufacturers, but also facilitates the development of more responsive and smarter home appliance systems through advanced FL and blockchain technologies.

Keywords

• Smarter home systems
• Federated learning
• Mobile edge computing
• Blockchain