Differential Privacy Based Federated Learning Techniques in IoMT: A Review

Abstract

The ever-expanding landscape of the Internet of Medical Things (IoMT) is increasingly reliant on Federated Learning (FL) to construct cooperative, privacy-centric AI models. By enabling model training on dispersed data sources, FL maintains the security of sensitive healthcare information while promoting the development of global models to augment the realm of medical care. To effectively mitigate privacy apprehensions intrinsic to healthcare data, the integration of differential privacy with FL emerges as a compelling strategy. This amalgamation not only offers robust privacy assurances but also facilitates the customization of model updates, ensuring the safeguarding of individual user data. This review aims to promote knowledge on the synergies between differential privacy and Federated Learning in IoMT. It is intended to benefit healthcare professionals, data scientists, policymakers, and technologists, by providing insights on privacy-preserving AI models, techniques to integrate FL and differential privacy, and designing secure and efficient IoMT solutions.