In order to use real-world health data for machine learning tasks, a number of practical issues must be resolved, including distributed data silos, privacy issues with compiling sensitive personal information into a single database, resource limitations for transferring and integrating data from various sites, and the possibility of a single point of failure. With the help of scattered health data held locally at numerous sites, this invention built a privacy-preserving federated learning (PPFL) architecture capable of learning a global model. Artificial intelligence (AI)-powered smart healthcare has changed as a result of recent advances in communication technologies. Due to the high scalability of contemporary healthcare networks and escalating data privacy concerns, practical healthcare contexts may not be able to accommodate AI approaches' traditional requirement for centralized data collection and processing. In order to complete AI training without the need for raw data exchange, federated learning (FL), a new distributed collaborative AI paradigm, coordinates several clients. This makes FL particularly interesting for smart healthcare. A privacy-preserving approach to FL model development is essential, especially in the healthcare industry where patient data is of the utmost importance. The importance of FL with Differential Privacy (DP) for smart healthcare is highlighted in this study. Since Federated Learning, unlike other approaches, has no impact on system speed, it is one of the most often used methods for training machine learning models.