Comparison of Machine Learning Techniques For Intrusion Detection System

No Thumbnail Available

Date

2018-10

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

The rapid growth in the ubiquity and sophistication of Information Communication Technology (ICT) and the emergence of new networking paradigms such as Cloud Computing (CC), and Internet of Things (IoT) have made vital changes in the globe. Computer network security is one of the most critical issue as attackers are also evolving dynamically. There should be a mechanism that fill the security vulnerability. One of the promising technique to ensure computer network security is the use of hybrid machine learning (ML) techniques which automate the process of intrusion detection in computer networks. In this research, six hybrid ML models were developed based on the Knowledge Discovery in Database (KDD) process model. The dataset used in this study has been taken from University of New Brunswick Institute (Canada Institute of Cyber Security). After selecting the dataset, preprocessing techniques such as filling missing records, reduce dimension, selecting the most relevant features, and finally normalize the dataset input using features scaling are performed. The hybrid ML models for intrusion detection systems (IDS) are implemented using Python programming language. In this work, a total of 274208 dataset records are used for the ML models evaluation. Out of this, 191945 datasets are used for training and a separate 82263 records are used as a testing set. The decision tree (DT) and neural network (NN) algorithms as supervised and K-means algorithm as unsupervised algorithms are applied in both without feature selection and with feature selection. The principal component analysis and decision tree (PCA-DT) model showed the best results in all performance parameters. The model has a prediction accuracy of 99.89% and the lowest false positive rate of 0.027%. Results confirm the effectiveness of our proposed methods.

Description

Keywords

Intrusion Detection, Machine learning algorithms, Computer Security

Citation