PRIVACY-AWARE FEDERATED GRAPH NEURAL NETWORKS FOR INTELLIGENT SMART CITY APPLICATIONS

Smart cities rely on interconnected systems that continuously generate large volumes of heterogeneous data from transportation networks, healthcare services, energy grids, surveillance systems, and environmental sensors. Centralized machine learning approaches often require collecting sensitive information into a single repository, increasing privacy risks and communication overhead. Federated Learning (FL) has emerged as a distributed learning paradigm that enables collaborative model training without exposing raw data. However, conventional federated learning methods struggle to capture complex relationships among interconnected smart city entities. This paper proposes a Privacy-Aware Federated Graph Neural Network (PA-FGNN) framework that combines Graph Neural Networks (GNNs), federated learning, and differential privacy to support secure and intelligent smart city applications. The framework learns graph-based representations from decentralized data while preserving user privacy through local model training and privacy-preserving parameter aggregation. Experimental evaluation demonstrates improvements in prediction accuracy, communication efficiency, and privacy protection compared with conventional centralized and federated approaches. The proposed framework provides an effective solution for scalable, secure, and intelligent smart city analytics.