Enhancing Intrusion Classification System Using Lite-CNN Modelling

Abstract

The increasing complexity and sophistication of cyber-attacks pose significant challenges to traditional intrusion detection systems, necessitating the development of more efficient and accurate classification models. This study proposes an enhanced intrusion classification system leveraging a lightweight Convolutional Neural Network (Lite-CNN) architecture to address the computational overhead and feature extraction limitations of conventional approaches. The proposed Lite-CNN model is evaluated using the CSE-CIC-IDS2018 dataset, which encompasses a diverse range of contemporary network traffic patterns and attack types, including DoS, DDoS, infiltration, brute force, and web-based attacks, ensuring comprehensive model validation. The architecture of Lite-CNN is optimized to extract relevant features with minimal parameters, enabling faster training and inference while maintaining high detection accuracy. Experimental results demonstrate that the proposed model achieves outstanding performance metrics, with a precision of 0.99, recall of 0.99, F1-score of 0.99, accuracy of 0.99, and Matthews Correlation Coefficient (MCC) of 0.98, outperforming several existing deep learning-based intrusion detection methods. These results highlight the model's ability to accurately distinguish between normal and malicious traffic with minimal false positives and negatives. The findings indicate that the Lite-CNN-based intrusion classification system is not only computationally efficient but also highly reliable, making it a promising solution for real-time network security monitoring. This research underscores the potential of lightweight deep learning models in cybersecurity, providing an effective framework for developing robust and scalable intrusion detection solutions capable of adapting to evolving cyber threats.