A Gamma-enhanced Naïve Bayes model for robust intrusion detection in IoMT networks

In the Internet of Medical Things (IoMT), intrusion detection systems (IDS) must be lightweight yet capable of modeling highly skewed, burst-like traffic produced by medical devices and real-time monitoring nodes. Traditional Gaussian Naïve Bayes often underperforms in this setting due to its assumption of symmetric feature distributions. This paper introduces NBGamma, a modified Naïve Bayes classifier that replaces the Gaussian likelihood with a Gamma-based likelihood formulation, enabling more accurate modeling of positively skewed traffic features commonly observed in IoMT networks. Unlike prior Naïve Bayes extensions, the proposed approach mathematically integrates the Gamma distribution into the likelihood computation and formally models the IDS task as a security-driven optimization problem. Experiments on the CSE-CIC-IDS2018 dataset demonstrate significant improvements in precision, recall, F1-score, and ROC-AUC compared to Gaussian NB under both unbalanced and SMOTE-balanced settings. These results highlight the ability of NBGamma to more effectively capture malicious traffic patterns, advancing lightweight IDS modeling for resource-constrained IoMT environments. Article Highlights Strengthens cyberattack detection in Internet of Medical Things healthcare environments. Detects abnormal behavior more accurately under both normal and highly imbalanced conditions. Delivers more stable and reliable protection than commonly used intrusion detection approaches.