Data Balancing and CNN based Network Intrusion Detection (KEY IJP************377)

• M.sharmila Devi

Abstract

The help of an automated process that filters and classifies network intrusions is often needed by cyber-security professionals. The classification of the attack type is essential for applying specific preventive measures to secure networks. Numerous Machine Learning (ML) models have been proposed as the foundation for Network Intrusion Detection (NID) systems. Yet, their efficacy varies based on many factors. For instance, an ML model trained on a highly unbalanced dataset may be biased towards over-represented attack types. On the other hand, focusing solely on the ML model's performance in minority classes can have a negative impact on its performance in the majority classes. We proposes a Network Intrusion Detection (NID) system that addresses the issue of imbalanced datasets and uses Convolutional Neural Networks (CNN) to classify different attack types. The performance of the proposed system is compared to other systems that use different techniques such as Random Over-Sampling (ROS), Synthetic Minority Oversampling Technique (SMOTE) and Adaptive Synthetic Sampling (ADASYN) for data balancing. The NSL-KDD and BoT-IoT datasets are used for benchmarking, and the results show that the proposed system performs well in the minority classes on the binary classification task. Our proposed system scores a good weighted average F1-Score on the multi-class classification task using the Bot-Iot dataset. etc.