Author(s): Dhyaa Hasan Khamees and Essa Ibrahim Essa

Abstract: The Internet of Things (IoT) is currently receiving significant attention from researchers and industries. Every day, smaller and smarter devices are being deployed across various IoT applications. However, protecting these devices from attacks is essential for their more efficient and effective operation. Security vulnerabilities often arise in IoT-based systems, disrupting their functionality and the services they provide. To address and mitigate these threats, intrusion detection systems (IDS) must be created that are efficient, effective and adapted to the continuing evolution of attack types and methods. With the great development the world is witnessing in artificial intelligence systems, exploiting the great ability of machine learning techniques to analyze and discover patterns within anomaly detection systems in Internet of Things networks has become crucial to enhancing the ability to detect threats and respond to them in a timely and effective manner. In response to these requirements, we are heavily involved in applying deep learning techniques to detect anomalies in IoT networks. The research aims to exploit the powerful analytical capabilities of machine learning and employ them in the process of improving the accuracy and efficiency of intrusion detection systems (IDS), which in turn leads to enhancing the ability of these systems against cyber threats. This research focuses on exploring the challenges and concerns associated with the Internet of Things (IoT) and proposing an Intrusion Detection System (IDS) model. In this paper, a hybrid approach is proposed to classify 9 types of attacks on IoT networks. Initially, both the SVM and Logistic Regression algorithms are used to extract the most important features from the data set, and then the Random Forest algorithm is applied to the selected features. The aim of this research is to reach an IDS model that is flexible and efficient while reducing the computational complexity of the random forest algorithm, which helps reduce the response time to threats that can be implemented on Internet of Things networks. The results obtained demonstrated that the SVM-RF model achieved the highest performance in detecting the type of attack with an accuracy of 95%. The research results reveal that applying the proposed hybrid model significantly enhances the effectiveness of IDS systems. The study shows the effectiveness of the proposed model in classifying types of attacks within IoT networks. The originality of this research lies in the innovative design of the proposed model and feature architecture that provides a robust solution to the evolving challenges of cybersecurity. This research contributes to this field by presenting an advanced and effective approach to detecting attacks within IoT networks, thus increasing the capabilities of security systems within these networks, which represent the basis of future technology and which will greatly impact various sectors and activities.

DOI: 10.33545/2707661X.2024.v5.i1b.82

Pages: 75-87 | Views: 85 | Downloads: 43

Download Full Article: Click Here