Comparison of Data Dimension Reduction Methods in The Problem of Detecting Attacks

Abstract

Data dimension reduction issue is an important problem in the data pre-processing stage of data intelligent computing systems. The performance of data dimension reduction methods not only ensure compatibility with machine learning techniques, but also improve data processing efficiency. However, the performance of a dimensional reduction processing method in a data set is always an open challenging issue since it is closely tied to the data features. This paper presents the results of comparing the performance of several approaches in two common approaches on the UNSW-NB 15 data set for attack detection. Our experimental results show that RF-MLP method is very effective for deploying IDSs against DOS attacks.