Abstract

In the ever-evolving landscape of IoT sensor networks, ensuring robust information security is imperative. This paper introduces a novel approach, the Temporal GAN Ensemble with Bagging (TGE-Bag), designed to fortify the security framework of IoT sensor networks. TGE-Bag leverages the power of Generative Adversarial Networks (GANs) with a temporal dimension, addressing the dynamic nature of IoT data streams. The ensemble aspect incorporates Bagging, enhancing the overall resilience and robustness of the security model. The temporal dimension in TGE-Bag recognizes the time-sensitive nature of IoT data, acknowledging that threats and anomalies may manifest differently over time. By incorporating GANs, the model can effectively generate synthetic data representative of the temporal patterns, allowing for more comprehensive training and robust anomaly detection. The ensemble approach further contributes to the model robustness by aggregating diverse GANs, each specialized in capturing specific temporal nuances. This paper evaluates TGE-Bag efficacy through extensive simulations on real-world IoT datasets, demonstrating its superior performance in detecting and mitigating security threats. The ensemble ability to generalize across diverse temporal patterns contributes to its adaptability in various IoT sensor network scenarios.