A Novel Decentralized Privacy Preserving Federated Learning Model for Healthcare Application

With the increasing usage of deep learning algorithms in various aspects of science, specifically in healthcare and disease diagnosis, challenges in data availability are encountered. In the healthcare industry, due to the sensitivity surrounding patient privacy, data collection is limited. To train a high-quality deep learning model capable of aiding in disease diagnosis, a large amount of data is required, which is not feasible under these conditions. One solution to utilize healthcare data for training a high-quality deep learning model is to use distributed deep learning models, where data is kept locally but still participates in the training process. Some implementations of distributed deep learning, such as federated learning exist, but these implementations have vulnerabilities in preserving data privacy. In this research, the goal is to develop a distributed deep learning model that defends user privacy against attacks from both inside and outside the network while maintaining the quality of the deep learning model.