Ebook: Federated Learning for Internet of Medical Things: Concepts, Paradigms, and Solutions

This book intends to present emerging Federated Learning (FL)-based architectures, frameworks, and models in Internet of Medical Things (IoMT) applications. It intends to build on the basics of the healthcare industry, the current data sharing requirements, and security and privacy issues in medical data sharing. Once IoMT is presented, the book shifts towards the proposal of privacy-preservation in IoMT, and explains how FL presents a viable solution to these challenges. The claims are supported through lucid illustrations, tables, and examples that present effective and secured FL schemes, simulations, and practical discussion on use-case scenarios in a simple manner. The book intends to create opportunities for healthcare communities to build effective FL solutions around the presented themes, and to support work in related areas that will benefit from reading the book. It also intends to present breakthroughs and foster innovation in FL-based research, specifically in the IoMT domain. The emphasis of this book is on understanding the contributions of IoMT to healthcare analytics, and its aim is to provide insights including evolution, research directions, challenges, and the way to empower healthcare services through Federated Learning.

Federated Learning (FL) is a newly introduced technology that has piqued the interest of researchers eager to investigate its potential and applicability. Federated learning simply tries to answer the question: “Can we prepare the model without trying to transfer data to a central location?”. Furthermore, federated learning allows for training without the need for data dissemination, which was not previously available with typical Machine Learning (ML) techniques. Google, Amazon, and Microsoft dominate the Artificial Intelligence (AI) market by providing cloud-based API and AI solutions. Traditional AI approaches provide confidential user data to servers whereby models are trained. Federated learning emerges from the confluence of on-device AI and ML, blockchain technology, and cloud technologies and Internet of Things (IoT).

Suppose that our centralized ML implementation will mean that all devices will have a local copy, which users will be able to use as needed. The model will now begin to learn and train itself based on the data available by the users, gradually growing wiser. The systems are then allowed to transfer training effectiveness from the local copy of the ML app to the central server. The same issue happens on many devices that have a local copy of the app. The findings will be collected on the centralized server, but this time without any user data. Nowadays, the number of modern devices is rapidly increasing, resulting in data being generated in large amounts by data generation. These devices are now used with numerous sensors to generate data, which is critical for client-users.

The book also intends to cover the ethical and social issues around the recent advancements in the field of decentralized Artificial Intelligence. The book is mainly intended for undergraduates, post-graduates, researchers, and healthcare professionals who wish to learn FL-based solutions right from scratch, and build practical FL solutions in different IoMT verticals.