Comparative Analysis of Traditional Machine Learning and Privacy Preserving Federated Learning

Abstract

With the rapid development in the IT field, thousands, even millions of IoT devices were developed. IoT devices play a vital role in the field of healthcare. Nowadays smart wearable devices are used in the field of healthcare to monitor the health of patients like heartbeat, fitness, blood pressure, etc. These IoT devices generate a vast variety of data, but in healthcare, the generated data is related to patients. This data contains the private and sensitive information of the patient. In real world, there are lot of big data generated on daily bases from different sources. These data are in hundreds of gigabytes, and it requires large storage devices. Artificial Intelligence and Machine learning is a technique that is used to predict the result on the base of given data. In machine learning, it requires the data to be present in a centralized location, which is a major security concern for the users. FL is a machine learning technique that trains algorithms across multiple decentralized devices. FL works on decentralized heterogeneous computing devices. It varies in many ways from traditional Machine Learning like time saving, resource saving, etc. FL is one of the types of machine learning that improve the privacy and security concerns. In FL one is a server that contains the main model. The server shares the model with clients and clients train the local model collaboratively on the bases of data. This technique protects the user from transferring data, and it minimizes privacy issues. For privacy preserving, we will use the Pailliar Homomorphic Encryption.