Healthcare data from clinical institution and pharmaceutical industries are fragmented and are sensitive with protected health information of individual. Data privacy, safety and security are of prime important. It is vital to train machine learning model without compromising privacy. In federated learning the global model is aggregated on aggregation servers as per the parameter of local model instead of local data. There is waiting for straggler client before aggregation in each round .Every client operated at its own speed of training. Asynchronous federated learning aggregates the global model based on available updated model received at the server asynchronously during each round of training. Asynchronous federated learning improves the efficiency, performance, privacy and security. In this paper, based on the Lung cancer dataset, the classification model is built. With NVIDIA flare, synchronous federated learning is carried out with one server and three clients for spliited data of lung cancer of Kaggle. The performance of synchronous federated learning is observed similar to the classical machine learning model. Modeling of asynchronous federated learning is carried out with Petri Net and analyzed for the reachability analysis with workcraft software. Substantial improvement in medical data privacy, security and safety is observed.

Federated Learning , Asynchronous federated learning, Machine Learning, Artificial Intelligence