Federated learning has been introduced to facilitate knowledge sharing across multiple smart end devices in a cloud environment. It serves as smart decision-making while protecting data privacy by avoiding raw data shared across end devices and a central server. Traditional Deep learning model incorporating federated learning to enhance the classification accuracy effectively by aggregating different sources of training samples.
   The existing federated learning approach with limited application-specific data, its learning efficiency, and accuracy become ineffective, and also it is crucial to address the challenge of knowledge sharing and data scarcity across different domains and heterogeneous applications. To tackle this constraint, federated transfer learning facilitates knowledge sharing and training with limited data for heterogeneous data applications. It incorporates the training process of new heterogeneous applications by utilizing the existing models with privacy preservation.