Resilient and sustainable smart cities: Cybersecurity, disaster management, and data-driven urban solutions

Abstract

Rapid urbanization has increased the complexity and fragility of modern cities, demanding more intelligent, data-driven, and resilient infrastructure systems. The concept of smart cities has emerged in response—emphasizing the integration of physical and digital layers, and enabling data-centric governance across sectors such as transportation, environment, disaster preparedness, and cybersecurity. This thesis is grounded in a five-layered smart city architecture, which conceptualizes urban systems across Physical, Communication, Data, Analytics, and Application layers, along with a Cross-Cutting layer. The study concentrates primarily on the Data and Analytics layers, and also incorporates the Cross-Cutting layer in certain use cases, developing artificial intelligence-based solutions to address critical challenges in diverse urban domains. While each solution addresses a distinct problem area individually, the proposed approaches are designed in a modular and generalizable manner, offering strong potential for future integration into more holistic and adaptive smart city systems. Five urban challenges are addressed, each corresponding to a dedicated area: Cybersecurity Management, Disaster Management, Air Pollution Forecasting, Traffic Prediction, and Tourism Demand Forecasting. In the cybersecurity domain, unsupervised, semi-supervised, and graph-based models are developed for anomaly detection for various network circumstances. In disaster management, graph representations of fault networks are used alongside deep learning to improve earthquake magnitude and time predictions. Air quality forecasting combines convolutional neural networks and long short term memory to capture both spatial features and temporal dependencies of pollution patterns. The traffic model leverages a city-scale graph dataset and graph-based embeddings to predict congestion with high accuracy. Lastly, the tourism model applies attention-enhanced deep learning techniques to improve tourism demand forecasts, supporting adaptive urban service management. Beyond their individual contributions, these solutions demonstrate how modular and scalable artificial intelligence frameworks can collectively strengthen urban resilience and sustainability. The thesis presents reproducible models, highlights opportunities for cross-domain integration, and offers a foundation for building smarter and more adaptive urban systems.