A Federated Learning Mechanism for Preserving Security of Sensitive Data

Federated Learning (FL) is a distributed machine learning(ML) method which allows for model training on information distributed across various servers and devices without sharing the sensitive data. It works by aggregating updates and weight values from the devices while keeping raw data private and secure. FL is a field which is widely used in finance, healthcare, and IoT systems along with personalized recommendations. Traditional ML and deep learning (DL) models are capable of producing good outcomes, however, these algorithms do not provide data security due to directly working on the raw data. Moreover, the FL can avoid security concerns due to its behaviour of sending the model to the data instead of working on the raw data itself. In this work, the FL has been introduced along with its three kinds of architectures. The motivation behind this work is that healthcare departments and financial institutes avoid sharing patients' sensitive information due to privacy concerns. However, the ML and DL models require a vast volume of data for accurate prediction outcomes. Hence, FL has been identified as the best model for working on sensitive data without privacy concerns. Furthermore, the work has also introduced the core challenges and application areas of FL. Moreover, in future, the FL architectures will be implemented with the healthcare dataset to increase data security and efficiently identify the outcomes.