Federated Learning in Secure Smart City Sensing
Monika Gandhi, Sushil Kumar Singh, R. N. Ravikumar, Krunal N. Vaghela · 2025
Smart cities utilize cutting-edge technologies to improve the quality of urban life. Integrating federated learning in sensor networks is crucial in developing smart cities. It facilitates the cooperative training of models across distributed devices, enhancing data privacy and diminishing the need for centralized data storage. This paper comprehensively examines the obstacles and potential advantages of integrating federated learning within smart city sensing systems. Collaborative model training in smart city sensing systems requires efficient and privacy-preserving methods due to the large volume of data generated. A promising paradigm known as Federated Learning (FL) has emerged, allowing for the decentralized training of models across a network of distributed edge devices. This study examines the various obstacles and potential advantages associated with adopting Federated Learning within the domain of smart cities and the distinct attributes of smart city environments, including various sensor types, different data modalities, and privacy concerns. This paper examines the technical obstacles associated with federated learning within the given context, encompassing issues such as communication overhead, strategies for model aggregation, and the ability to adapt to dynamic urban environments. In addition, we emphasize the potential for augmenting smart city applications using Federated Learning (FL), including advancements in model generalization, decentralized decision-making, and heightened privacy preservation. By thoroughly examining existing literature and case studies, we offer valuable insights into the present condition of Federated Learning (FL) in the context of intelligent city sensing. In conclusion, we present a comprehensive overview of potential avenues for research and approaches to tackle the aforementioned obstacles, thereby promoting federated learning as a resilient and expandable remedy for advancing smart city sensing technologies.