Analysis on the Impact of Feature Selection on Cloud Intrusion Detection

Abstract

With the rapid development of cloud computing, the security of cloud system is getting more and more attention. The intrusion detection technology based on machine learning plays a huge advantage in protecting cloud system, but the data involved in cloud system is often large in scale, high in dimension, and strong in redundancy, which will greatly affect the detection performance and efficiency of machine learning models in cloud system. As one of the perfect solutions, feature selection has been widely used, but there is still a lack of quantitative analysis of the impact of overall feature selection methods on the performance of intrusion detection. Therefore, this paper systematically studies the impact of different feature selection methods on cloud intrusion detection, including ANDVA, Chi-square test, Mutual Information, ExtraTree, Random Forest, Recursive Feature Elimination. Extensive experiments are carried out, and the experimental results show that, except SVM, the detection performance and detection efficiency of other machine learning models are greatly improved after applying different feature selection methods, and different machine learning models have different fitness for different feature selection methods. However, SVM has certain differences in detection performance after applying different feature selection methods.