AI-Driven IoT Security Framework for Real-Time Threat Detection in Large-Scale Smart City Deployments

The goal of this research is to develop and deploy an AI-enabled IoT security framework for large smart city environments that supports real-time threat detection and adaptive response. The rapid proliferation of IoT devices in smart cities has magnified security risks due to the interconnected nature of the underlying structures. The old model of signature-based intrusion detection is fundamentally reactive and thus suffers from delayed response to the new strategies employed by attackers. The situation calls for an automated, intelligent, quick response model which is robustly scalable. This research presents an architecture that real-time threat analytics using edge computing in a federated learning environment with multiple integrated convolutional and long short-term memory neural networks. The system performs anomaly-based detection using data preprocessing, feature selection, federated model training and privacy-preserving model aggregation to enhance collusion privacy and model accuracy. Model validation was done using a hybrid dataset consisting of CICIDS2018 and IoT-23. Experimental results show the framework achieving 97.8% detection accuracy, a 0.021 false positive rate, and 25% less processing latency to centralized IDS models. The proposed framework demonstrates low latency, and high scalability and adaptability to zero-day attacks, providing a practical framework to build from for smart cities. Future work will incorporate blockchain for decentralized trust management.

Ҳали таржима қилинмаган