Smart cities are expeditiously incorporated in many underdeveloped and developed nations as a consequence of evolution in cloud services, big data, Artificial Intelligence (AI), the Internet of Things (IoT), and fifth-generation (5G) technologies. The smart city comprises several networking and computing systems to expedite the overall quality of life for its citizens. Data privacy and security are essential to implementing a feasible smart city. Conventional machine learning techniques to develop smart cities experience impediments with data privacy and security due to the centralization of data from different sources in one place, which are ease of threats.
Federated learning is a promising technique in preserving data privacy for distributed data. Federated learning offers collaborative construction of smart city technologies through training the learning model locally and aggregates the local updates without sharing the sensitive data of the users. Application domains of federated learning via smart city are IoT system - edge computing, block-chain, transportation system - vehicle communication, electric vehicle, autonomous vehicle, aviation system - aircraft, Unmanned Aerial Vehicle (UAV), finance - financial fraud, insurance, medical field - medical data privacy protection, drug development, disease prediction and in communication - wireless multiple-access channel. Future directions of federated learning in smart cities are the defense against attacks and effective optimization algorithms.