Federated learning for secure and privacy-preserving edge AI in smart cities’ wastewater management systems

Abstract

The rapid evolution of smart cities has significantly transformed wastewater management systems through the integration of real-time monitoring, intelligent control, and data-driven decision-making. However, conventional centralized artificial intelligence (AI) frameworks face critical limitations, including data privacy vulnerabilities, security risks, high communication overhead, and poor scalability across geographically distributed treatment facilities. This study investigates federated learning (FL) as an enabling paradigm for developing secure, privacy-preserving, and scalable edge AI solutions tailored to smart wastewater infrastructure. Unlike traditional centralized approaches, FL facilitates decentralized model training across wastewater treatment plants by sharing model parameters instead of raw data, thereby reducing data transmission costs and ensuring compliance with data protection regulations. To enhance system robustness, a hybrid framework is proposed that integrates differential privacy, secure multi-party computation (SMPC), and blockchain-based authentication. These mechanisms collectively strengthen resilience against adversarial attacks, prevent data leakage, and ensure the integrity of model updates. Furthermore, an adaptive aggregation strategy is introduced to address data heterogeneity across distributed facilities by dynamically weighting model contributions based on sensor reliability, data quality, and network conditions. The proposed framework is evaluated using representative wastewater datasets, demonstrating significant improvements in anomaly detection accuracy, robustness to non-identically distributed data, and privacy preservation compared to conventional centralized AI models. Additionally, the framework achieves enhanced energy efficiency and reduced communication overhead in edge environments. The findings establish FL as a foundational technology for deploying secure, scalable, and privacy-aware AI systems in smart wastewater management.