article · Journal of Communications Software and Systems
The growth of the Internet of Things has created significant security vulnerabilities, as connected devices can be compromised to steal sensitive data or launch major cyberattacks. This research examines four deep learning architectures for detecting and categorising intrusions: a one-dimensional convolutional neural network, a long short-term memory model, a hybrid combining both, and a two-dimensional convolutional neural network. The systems were evaluated using the BoTNeTIoT-L01-v2 dataset, which comprises normal network traffic alongside attack data from various connected devices. The workflow involved data preprocessing, feature extraction, and model training, including weighted variants designed to optimise feature importance. The two-dimensional convolutional network and the hybrid architecture delivered the best performance, recording high classification accuracy. These findings offer comparative insights into how weighted features can improve threat identification for future monitoring systems.
Connected devices are increasingly embedded in daily life and critical infrastructure, yet they remain frequent targets for cyberattacks. Establishing which machine learning techniques best identify malicious traffic helps security developers build more reliable automated defences, protecting private information and preventing networks of compromised hardware from disrupting broader digital services.
This work informs developers of network security tools and intrusion detection systems seeking to enhance automated threat recognition in connected device environments. Because the findings are based on benchmarking against an offline dataset, the research represents early-stage development that requires testing in live network environments before real-time commercial deployment.
AI-generated from the published abstract. Always read the original work before citing.
The Internet of Things (IoT) has provided many benefits to society and introduced new security challenges. Attackers can target IoT devices to steal sensitive information or launch large-scale attacks. In this field, deep learning algorithms have provided encouraging results in the discovery and classification of intrusions in IoT devices. This study investigates the implementation and performance of four deep learning models: One-Dimensional Convolutional Neural Network (1DCNN), Long Short-Term Memory (LSTM), a hybrid 1DCNN-LSTM, and Two- Dimensional Convolutional Neural Network (2DCNN) for detecting and classifying IoT device attacks. Using the BoTNeTIoT-L01- v2 dataset, which includes normal and attack traffic provided by various IoT devices, we preprocess the data, extract features, and train the models, including weighted versions to optimize feature importance. Our findings highlight that the 2DCNN and hybrid 1DCNN-LSTM models shows superior performance, achieving high classification accuracy. This study contributes a comprehensive comparative analysis of deep learning models for IoT security, focusing on the effectiveness of weighted features in improving detection accuracy. The results provide valuable information for the advancement of real-time IoT attack detection systems.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.24138/jcomss-2024-0101
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
