An Effective RF-based Intrusion Detection Algorithm with Feature Reduction and Transformation

Abstract

Intrusion detection systems are essential in the field of network security. To improve the performance of detection model, many machine learning algorithms have been applied to intrusion detection models. Higher-quality data is critical to the accuracy of detection model and could greatly improve the performance. In this paper, an effective random forest-based intrusion detection algorithm with feature reduction and transformation is proposed. Specifically, we implement the correlation analysis and logarithm marginal density ratio to reduce and strengthen the original features respectively, which can greatly improve accuracy rate of classifier. The proposed classification system was deployed on NSL-KDD dataset. The experimental results show that this paper achieves better results than other related methods in terms of false alarm rate, accuracy, detection rate and running time.