A Novel Generative Adversarial Network Approach to Autonomous Threat Detection and Prevention in Educational Internet of Things Platforms

Abstract

Educational Internet of Things (IoT) systems generate sensitive student data through real-time learning analytics, monitoring, tracking, and adaptive instruction, creating security challenges for conventional intrusion detection systems that rely on labeled attack data and exhibit high false-positive rates. This study developed and evaluated a lightweight, edge-deployable framework based on a Wasserstein generative adversarial network with gradient penalty (WGAN-GP) for autonomous threat detection and proactive prevention in educational IoT environments. The generator produced synthetic attack patterns to address extreme class imbalance, while the discriminator learned normal student-device behavior and distinguished anomalies in real time. A policy engine operating on the same edge node autonomously triggered alert escalation, device isolation, and traffic rerouting upon threat detection. The framework was implemented on an ESP32-based smart-classroom testbed using augmented CIC-IoT 2023 data and MQTT-based educational traffic traces. It achieved 98.7% detection accuracy, an average inference latency of 0.12 s, and a 94% reduction in false positives compared with conventional machine learning baselines. Its reported memory footprint was below 180 [unit to be confirmed]. These findings indicate the framework’s potential to integrate real-time threat detection with automated preventive responses at the edge. The study contributes a generative adversarial network framework for autonomous defense in educational IoT environments, with practical implications for protecting sensitive student data.