Scalable Intrusion Detection in IoT Networks Via Property Testing and Federated Edge AI

The rapid proliferation of Internet of Things (IoT) devices has expanded the attack surface for cyber threats, necessitating real-time and scalable security solutions. Traditional Intrusion Detection Systems (IDS) require extensive data processing, which is computationally infeasible for resource-constrained IoT environments. This paper proposes a novel property testing-based anomaly detection framework that efficiently monitors IoT networks while reducing computational overhead. By leveraging sublinear-time statistical sampling and probabilistic validation, the proposed framework detects network anomalies, unauthorized access, and malware propagation without requiring full dataset analysis. A hybrid Edge AI and Federated Learning model is integrated to enhance security across distributed IoT nodes while ensuring data privacy. The system is evaluated against real-world IoT cybersecurity threats, demonstrating up to 70% reduction in processing time and a 92% accuracy rate in detecting intrusions with only 10% of the dataset analyzed. Compared to recent lightweight federated and sketch-based IDS models, the proposed framework improves F1-score by up to 11.3%, while reducing inference time and energy consumption by over 60%, highlighting its practical efficiency and scalability. These findings underscore the potential of property testing as an efficient alternative for real-time cybersecurity monitoring in IoT ecosystems.