An Empirical Study on Imbalanced Learning in Intrusion Detection Using Random Tree Classifier
Ritinder Kaur, Neha Gupta · 2022 International Conference on Augmented Intelligence and Sustainable Systems (ICAISS) · 2022
Intrusion detection system (IDS) plays a crucial role in the field of network security. With the burgeoning surge of network users and smart devices, new risks keep arising which require regular updating of security mechanisms. Machine learning is one of the promising methods in designing IDS as it offers low false alarm rate and high detection rate. However, network data is inherently imbalanced in nature with skewed distribution. Conventional learners tend to be biased towards the majority normal class and hence misclassify the minority attack classes which can be detrimental to the network. This paper discusses the different approaches available for improving the class imbalance and provides an empirical comparison between these approaches on a widely used intrusion detection dataset NSL-KDD. The result shows that by using a befitting imbalanced learning approach and feature selection strategy, the identification of rare class attacks can be substantially improved.