[Paper Review] Federated Learning in Smart Cities: A Comprehensive Survey.

Zhaohua Zheng, Yize Zhou+4 more|arXiv (Cornell University)|Feb 2, 2021

Privacy-Preserving Technologies in Data5 citations

TL;DR

This paper provides a comprehensive survey of federated learning (FL) in smart cities, highlighting its role in enabling privacy-preserving AI across domains like IoT, transportation, healthcare, and finance. By reviewing key FL technologies, architectures, and recent advances, the study identifies critical challenges and outlines future research directions for scalable, secure, and efficient FL deployment in urban environments.

ABSTRACT

Federated learning plays an important role in the process of smart cities. With the development of big data and artificial intelligence, there is a problem of data privacy protection in this process. Federated learning is capable of solving this problem. This paper starts with the current developments of federated learning and its applications in various fields. We conduct a comprehensive investigation. This paper summarize the latest research on the application of federated learning in various fields of smart cities. In-depth understanding of the current development of federated learning from the Internet of Things, transportation, communications, finance, medical and other fields. Before that, we introduce the background, definition and key technologies of federated learning. Further more, we review the key technologies and the latest results. Finally, we discuss the future applications and research directions of federated learning in smart cities.

Motivation & Objective

• To examine the current state of federated learning (FL) and its integration into smart city infrastructures.
• To identify and analyze key FL technologies enabling data privacy while maintaining model performance across urban applications.
• To investigate FL's practical deployment in domains such as transportation, healthcare, finance, and communications within smart cities.
• To review recent advancements in FL algorithms, system design, and security mechanisms tailored for urban-scale data processing.
• To outline open challenges and future research directions for scalable, efficient, and privacy-preserving FL in smart city ecosystems.

Proposed method

• Conduct a systematic literature review of federated learning applications in smart city domains including IoT, transportation, and healthcare.
• Categorize and analyze key FL technologies such as horizontal, vertical, and transfer FL, focusing on their suitability for urban data environments.
• Examine core components of FL systems, including client selection, model aggregation (e.g., FedAvg), and communication efficiency techniques.
• Review security and privacy mechanisms like differential privacy and secure aggregation used in FL to protect sensitive urban data.
• Evaluate system-level challenges such as non-IID data, device heterogeneity, and communication bottlenecks in smart city deployments.
• Synthesize findings into a structured overview of FL’s technical foundations, application areas, and deployment considerations in urban settings.

Experimental results

Research questions

• RQ1 How does federated learning enable privacy-preserving AI in smart city applications?
• RQ2 What are the key technical components and system architectures that support federated learning in urban environments?
• RQ3 In which smart city domains—such as transportation, healthcare, and finance—has federated learning been successfully applied?
• RQ4 What are the main challenges in deploying federated learning at scale in smart cities, including data heterogeneity and communication efficiency?
• RQ5 What future research directions are critical for advancing federated learning in smart city ecosystems?

Key findings

• Federated learning effectively addresses data privacy concerns in smart cities by enabling model training without centralizing raw data.
• The integration of FL in IoT and transportation systems enables real-time, privacy-preserving decision-making using decentralized data sources.
• Key technologies such as FedAvg and secure aggregation significantly improve model convergence and data security in urban AI applications.
• Despite progress, challenges like non-IID data distributions and device heterogeneity remain significant barriers to widespread FL deployment in smart cities.
• Future research must focus on improving communication efficiency, robustness to system failures, and scalability across heterogeneous urban infrastructures.
• The combination of federated learning with edge computing and blockchain shows promise for enhancing trust and performance in smart city systems.