An Effective Feature Selection Model Using Hybrid Metaheuristic Algorithms for IoT Intrusion Detection.

Saif S Kareem,Fatma A Hashim,Reham R Mostafa,Hazem M El-Bakry

doi:10.3390/s22041396

Saif S Kareem, Fatma A Hashim + Show 2 more

Open Access

https://doi.org/10.3390/s22041396

Copy DOI

Abstract

The increasing use of Internet of Things (IoT) applications in various aspects of our lives has created a huge amount of data. IoT applications often require the presence of many technologies such as cloud computing and fog computing, which have led to serious challenges to security. As a result of the use of these technologies, cyberattacks are also on the rise because current security methods are ineffective. Several artificial intelligence (AI)-based security solutions have been presented in recent years, including intrusion detection systems (IDS). Feature selection (FS) approaches are required for the development of intelligent analytic tools that need data pretreatment and machine-learning algorithm-performance enhancement. By reducing the number of selected features, FS aims to improve classification accuracy. This article presents a new FS method through boosting the performance of Gorilla Troops Optimizer (GTO) based on the algorithm for bird swarms (BSA). This BSA is used to boost performance exploitation of GTO in the newly developed GTO-BSA because it has a strong ability to find feasible regions with optimal solutions. As a result, the quality of the final output will increase, improving convergence. GTO-BSA’s performance was evaluated using a variety of performance measures on four IoT-IDS datasets: NSL-KDD, CICIDS-2017, UNSW-NB15 and BoT-IoT. The results were compared to those of the original GTO, BSA, and several state-of-the-art techniques in the literature. According to the findings of the experiments, GTO-BSA had a better convergence rate and higher-quality solutions.

Highlights

Accepted: 7 February 2022The Internet of Things (IoT) has emerged in the modern era, pushing the development of new business process technologies through a network of computers and devices capable of communicating and engaging with one another [1]
To enhance the global searching and local searching abilities, we propose a novel improvement of the Gorilla Troops Optimizer (GTO) algorithm using four strategies: (1) the control randomization (CR) parameter and (2) an advanced nonlinear transfer function to balance exploration and exploitation, (3) different settings in the GTO exploration phase, and (4) a novel local updating position strategy based on the bird swarms (BSA) algorithm
The specific nature of the Internet of Things (IoT) applications, which consist of millions of sensors, leads to generating a massive amount of data

Summary

Introduction

Accepted: 7 February 2022The Internet of Things (IoT) has emerged in the modern era, pushing the development of new business process technologies through a network of computers and devices capable of communicating and engaging with one another [1]. It is possible to characterize cloud computing as a model in which a variety of services and resources are made available to customers on demand, with little involvement from either the service provider or the customer [3]. Most IoT applications in different fields depend on cloud computing to store and process data. Cyberattacks on cloud computing have increased for several reasons, including the availability and accessibility of hacking tools, which led to the hacker not needing extensive knowledge or exceptional skills to carry out an attack [4]. The classification (i.e., supervised learning) of a dataset has NS × NF dimensions, where NS represents the total number of samples and NF denotes the number of features. The primary goal of the FS algorithm is to select a subset S from the total number of features (NF ), where the dimension of S is less than NF.

Objectives

Results

Conclusion