Intrusion Detection System based on Hybrid Feature Selection and Support Vector Machine (HFS-SVM)

摘要

In recent years, anomaly-based intrusion detection techniques are continuously developed and the Support Vector Machine (SVM) is one of the techniques. However, it requires training time and storage if there are lots of numbers of features. In this paper, a hybrid feature selection, using correlation-based feature selection and the motif discovery using random projection techniques, is proposed to reduce the number of features from 41 to 3 features with the KDD'99 dataset. It is compared with a regular SVM technique with 41 features. The results show that the accuracy rate is high at 98% and the training time is less than that of the regular SVM almost by half.