A CNN-Based Network Profiling System for Intrusion Detection: Addressing Evolving Cyber Threats

Abstract

Intrusion detection is a critical aspect of securing computer networks against unauthorized access and malicious activities. Traditional intrusion detection systems (IDS) are typically signature-based, which limits their ability to withstand evolving cyber threats. To address these limitations, this study proposes a novel approach: a Convolutional Neural Network (CNN)-based network profiling system for intrusion detection. The primary aim of this study is to design and develop a CNN-based network profiling system that enhances the accuracy and effectiveness of identifying and mitigating cyber threats in computer networks. The proposed system leverages a CNN architecture to extract high-level features from network packets, capturing intricate spatial dependencies within the data. This approach enables the network profiling system to efficiently distinguish between normal and anomalous traffic, making it highly adaptive to new and previously unseen attack vectors. Data was collected from diverse network environments, encompassing legitimate network behavior and various intrusion scenarios. The collected data was pre-processed and converted into suitable input representations for the CNN model. The network profiling system was subsequently trained on a large-scale dataset to enable it to learn complex patterns and anomalies present in network traffic. The evaluation was conducted through extensive experimentation using benchmark datasets and real-world traffic traces. A comparative analysis with traditional IDS methods demonstrates the superiority of the CNN-based approach in terms of accuracy, efficiency, and adaptability. This work highlights the scalability of the proposed system, ensuring its applicability to large-scale enterprise networks.