An SDN-AI-Based Approach for Detecting Anomalies in Imbalance Data Within a Network of Smart Medical Devices

The Internet of Medical Things (IoMT) has become a novel paradigm for real-time healthcare applications. Artificial intelligence (AI) based efforts have been made to address the security challenges of IoMT, the problem of imbalance data still exists, due to which AI algorithms cannot sufficiently learn malicious traffic behavior and fail to identify rare anomalies in imbalance data accurately. Therefore, in this article, we propose an intelligent model based on software defined networking and deep learning (DL) to handle the heterogeneous, complex, and distributed architecture. To tackle the imbalance challenge, the proposed model utilizes generative adversarial network (GAN) to generate plausible synthetic data for minor class traffic. It combines autoencoder-driven DL models with reconstruction error and Wasserstein distance-based GAN. When compared to naive and advanced techniques, the proposed model produced noticeably better results on an imbalance dataset and outperformed these techniques by 4.78% and 4.54% in terms of accuracy and F1-score values, respectively.