Dip-DARK: A smart and innovative classifier for enhanced intrusion detection and security in heterogeneous IoT networks

Abstract

Presently, significant research works focused on the design and development of security methods for protecting Heterogeneous Internet of Things (HetIoT) networks. Yet, the conventional approaches suffer with the problems of high processing time, lower accuracy, increased system designing complexity, and reduced efficiency. Therefore, in the proposed study, a novel and unique framework known as Dip-DARK—Dipper Throated Optimization integrated Deep Activation based Runge Kutta Classifier—is developed to safeguard the HetIoT network from potentially dangerous intrusions. Some of the well-known and most recent intrusion datasets, including as CIC-DDoS 2019, ToN-IoT, Edge-IIoT, and In-SDN, have been used for system development and validation. The proposed model is validated and tested by using these datasets. Then, to effectively shrink the dataset, the most important features are best selected using the Dipper Throated Optimization (DipTO) model, an intelligent optimization method. As a result, the Deep Activation based Runge Kutta (DARK) classifier was able to precisely predict the type of intrusion using the set of optimized features. Additionally, using a variety of performance measures, the proposed Dip-DARK model’s intrusion detection findings are evaluated and contrasted with current state-of-the-art model methodologies.