Blockchain-Assisted Federated Learning Framework for Data Processing in IoMT Domain

Federated Learning(FL) has been on the rise for the last few years. Its integration with devices with low processing power, especially where privacy is the main concern, has risen accordingly. In the case of the Internet of Medical Things (IoMT), where the data is private and must be protected, FL is the way to go. To add even a more layer of security, blockchain is used along with FL. The blockchain is stored in the cloud. but to maintain a lesser traffic propagation to the cloud and manage the data properly, we use a fog layer. In FL, all the devices in the IoMT environment contain low computation power. Hence, they use a simple, lightweight neural network for simulation. In this paper, we have used PathMNIST from the MedMNIST dataset to simulate FL for 10 rounds for 10, 20, 30, and 50 clients, respectively, and plotted accuracy and loss graphs. The simulation has been done in a single server setup using horizontal federated learning with the FedAvg strategy.