Enhancing IoMT Intrusion Detection through Nature Inspired Optimization-Based Gated Recurrent Unit Neural Network

Dash, Pandit Byomakesha and Nayak, Janmenjoy and Vimal, S. and Shah, Sayed Chhattan and Bilal, Muhammad (2026) Enhancing IoMT Intrusion Detection through Nature Inspired Optimization-Based Gated Recurrent Unit Neural Network. IEEE Open Journal of the Communications Society, 7. pp. 4129-4146. ISSN 2644-125X

Abstract

Internet of Medical Things (IoMT) devices have grown rapidly in size across Industrial Control Systems (ICS) and IoT-enabled Cyber-Physical Systems (CPS), thereby increasing susceptibility to more advanced cyberattacks, and the need to use highly sensitive and timely intrusion detection mechanisms. In this paper, a new GOOSE-optimized Gated Recurrent Unit (GOOSE_GRU) model has been suggested to make use of a powerful and intelligent intrusion-detecting model. The GOOSE metaheuristic algorithm is doing the most effective hyper-parameter optimization of the GRU framework and increasing convergence speed and classification stability. As demonstrated in experimental results, the proposed GOOSE_GRU model results in 100 percent accuracy, 1.0 ROC-AUC alongside almost perfect precision (0.9999), recall (0.9999) and F1-score (0.9999) on the test data, compared to conventional deep learning (DL) models such as DNN, RNN, CNN, LSTM, standard GRU, and other optimized variants such as PSO_GRU, GWO_GRU and FA_GRU which resulted in maximum accuracies Moreover, the suggested model lowers the computation cost at the expense of maintaining a high-quality generalization at the 10-fold stratified validation. These findings approve the efficacy, strength, and realistic acceptability of the suggested framework in safeguarding real-time IoMT-enabled ICS and CPS environments in regard to the emerging cyber dangers.