Addressing Imbalanced Classes Problem of Intrusion Detection System Using Weighted Extreme Learning Machine

Abstract

The main issues of the Intrusion Detection Systems (IDS) are in the sensitivity of these systems toward the errors, the inconsistent and inequitable ways in which the evaluation processes of these systems were often performed. Most of the previous efforts concerned with improving the overall accuracy of these models via increasing the detection rate and decreasing the false alarm which is an important issue. Machine Learning (ML) algorithms can classify all or most of the records of the minor classes to one of the main classes with negligible impact on performance. The riskiness of the threats caused by the small classes and the shortcoming of the previous efforts were used to address this issue, in addition to the need for improving the performance of the IDSs were the motivations for this work. In this paper, stratified sampling method and different cost-function schemes were consolidated with Extreme Learning Machine (ELM) method with Kernels, Activation Functions to build competitive ID solutions that improved the performance of these systems and reduced the occurrence of the accuracy paradox problem. The main experiments were performed using the UNB ISCX2012 dataset. The experimental results of the UNB ISCX2012 dataset showed that ELM models with polynomial function outperform other models in overall accuracy, recall, and F-score. Also, it competed with traditional model in Normal, DoS and SSH classes.