Skip to content
Network Intrusion Detection Using Deep Belief Network (DBN)

Network Intrusion Detection Using Deep Belief Network (DBN)

Ieeexplore.Ieee • November 4, 2026

The escalating Interconnectedness of devices has given rise to a surge in cyber-security threats, particularly zero-day attacks, which pose a significant challenge to tra...

The escalating Interconnectedness of devices has given rise to a surge in cyber-security threats, particularly zero-day attacks, which pose a significant challenge to traditional Intrusion Detection Systems (IDSs). This paper delves into the realm of behavior-based IDSs, leveraging Deep Neural Networks (DNNs), with a specific focus on the detection of network attacks. The effectiveness of these systems is closely tied to the quality of the training Dataset, where underrepresented samples can compromise detection performance, Our research centers on the development and evaluation of Deep Belief Networks (DBNs) as a robust solution for detecting cyber-attacks within networks of connected devices. The cornerstone of our investigation is the utilization of the CICIDS2017 Datasets, a comprehensive repository that facilitates both training and evaluation of our proposed DBN approach. To address the challenges posed by imbalanced Datasets, we employ various class balancing techniques, systematically evaluating their impact on detection performance. In the context of network attacks, our study pays special attention to notorious threats such as SQL injection, Cross-Site Scripting (XSS), Denial of Services (DoS), Distributed Denial of Services (DDoS), and other malicious activities. These attacks, often veiled in the subtleties of network traffic, demand sophisticated detection mechanisms. The empirical evaluation involves a comparative analysis with a conventional Multi-Layer Perceptron (MLP) model and the prevailing state-of-the-art IDSs. Our proposed DBN approach demonstrates competitive and promising results, showcasing significant performance improvements in the detection of attacks that are underrepresented in the training Datasets. The findings underscore the efficacy of DBNs in tackling the intricate challenges posed by emerging and sophisticated cyber threats. This research not only contributes to the advancement of intrusion detection methodologies but also provide...

2024 Intelligent Methods, Systems, and Applications (IMSA)

The ever-expanding geography of connected devices has steered in a period of unequaled convenience and effectiveness. still, this connected web has also given rise to a redoubtable array of cyber-security challenges, with zero-day attacks arising as pervasive trouble. Traditional Intrusion Discovery Systems(IDSs), counting on Deep Neural Networks(DNNs), stand at the van of defense against these sophisticated attacks. The efficacy of these systems, crucially dependent on the quality of their training Datasets, faces a redoubtable challenge in the form of underrepresented samples, which can compromise discovery performance.

References is not available for this document.