Federated Learning for Data Security and Privacy Protection

Abstract

In recent years, Artificial intelligence (AI) have been applied in many fields, including driverless cars, smart cities, healthcare, finance, etc. However, data island and data privacy security are still two major challenges for AI, federated learning is proposed as a solution. In federated learning, many clients collaborate to train a common model under the coordination of a central server, while keeping the training data decentralized. Each client's data does not leave the local area, and a global shared model is jointly built by means of parameter exchange under the encryption mechanism, the built model serves only the local target in their respective regions. In this paper, we present the definition, classification, and learning process of the federated learning, and discuss the key challenges faced by federated learning and the solutions that are currently available.