Contraction control factor-based gorilla troop optimizer for features in intrusion detection systems

Abstract

Internet of things (IoT) has evolved into a large-scale network due to the increasing number of connected devices and massive amount of data they generate. IoT networks produce massive amounts of heterogeneous data from various devices, making it difficult to identify relevant features for intrusion detection. Hence, this research proposes the contraction control factor-based gorilla troop optimizer (CCF-GTO) for feature selection and multiple parametric exponential linear units based long short-term memory (MPELU-LSTM) approach for classification of intrusion detection system (IDS) in IoT. CCF-GTO. It uses adjustable parameters to prioritize relevant information while eliminating unnecessary features, making the model more efficient and resulting in better classification accuracy. The experimental results demonstrate that the MPELU-LSTM approach achieves better accuracy of 99.56% on the UNSW-NB15 dataset as compared to the earlier approaches like convolutional neural network with LSTM (CNN-LSTM) and optimized deep residual convolutional neural networks (DCRNN). These findings suggest that the MPELU-LSTM method significantly enhances the accuracy and robustness of IDS in IoT environments by addressing issues like the identification of relevant features and feature redundancy, contributing to more effective and secure systems. This research has valuable implications for enhancing the security bearing of IoT infrastructure.