Hybrid Convolutional Neural Network-Based Intrusion Detection System for Secure IoT Networks

Security vulnerabilities are a growing concern due to the increasing prevalence of Internet of Things (IoT) devices. This paper presents a hybrid Convolutional neural network (CNN)-based intrusion detection system (IDS) for IoT networks that detects threats. The research tackles the shortcomings of...

Full description

Saved in:
Bibliographic Details
Main Authors: Sami Qawasmeh, Ahmad Habboush, Bassam Elzaghmouri, Qasem Kharma, Da'ad Albalawneh
Format: Article
Language:English
Published: Tikrit University 2025-08-01
Series:Tikrit Journal of Engineering Sciences
Subjects:
Online Access:https://tj-es.com/ojs/index.php/tjes/article/view/2526
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Security vulnerabilities are a growing concern due to the increasing prevalence of Internet of Things (IoT) devices. This paper presents a hybrid Convolutional neural network (CNN)-based intrusion detection system (IDS) for IoT networks that detects threats. The research tackles the shortcomings of existing IDSs, which focus on individual threats and are computationally expensive. The proposed method outperforms the traditional machine learning and deep learning models in identifying IoT network attacks. The goal of this study is to create an effective IDS for IoT networks that can detect a range of anomalies and malicious attacks. The research aims to address the limitations of existing intrusion detection systems (IDSs) by enhancing their ability to detect a broader range of attacks with improved performance through the addition of a "long short-term memory (LSTM)" component and the utilization of a hybrid CNN model. The proposed model includes data gathering, preprocessing, network training, and attack identification. System logs and their features are selected for data collection, followed by preprocessing to remove noise. The training model defines the convolutional layer's structure, sliding window size, neuron connection weights, and outputs using the improved data. The training period is used for attack detection, and the weights are calculated from trained and real-time data. Using the UNSW NB15 dataset, the suggested IDS is tested against a recurrent neural network (RNN) system. The suggested model outperforms the RNN model in several performance parameters, achieving a 99.1% detection accuracy, 4% higher. CNN-based intrusion detection in IoT networks is stressed in the study. It shows how hybrid CNN-based techniques can improve IoT network security and resilience. The proposed IDS introduces a novel approach that utilizes a hybrid CNN model and incorporates LSTM to enhance the detection capabilities of IoT network attacks. The study highlights the significance of leveraging advanced machine learning techniques to maintain the integrity and privacy of IoT systems.
ISSN:1813-162X
2312-7589