Federated Learning and Artificial Intelligence in IoT: Security Challenges, Solutions, and Future Directions

Abstract

There has been a great proliferation of the Internet of Things in different areas, ranging from healthcare, smart cities, industry, transport to agriculture, among others, and as a result of which an immense volume of distributed data is generated. The problem with traditional centralized machine learning methods is in the need to transfer sensitive data to centralized servers, which leads to additional risks of privacy breaches, communication overhead, and failure of single point. As a way of solving mentioned challenges, there appears the federated learning framework, which allows making collaborative model training without data transfer thus increasing the privacy of the system and saving on communications.