An Intelligent Intrusion Detection Scheme Powered by Boosting Algorithm

Abstract

Security is always a major concern in today’s world. Due to the prevalent techniques like Internet of Things (IoT), Fog/Edge computing and the vast use of social networking, there is a significant increase in the generation of network traffic data. For this reason, proper and fast mechanisms are needed to monitorvariety of data to fight against the vulnerabilities and threats those may occur in the system. In the present article, a machine learning based Intrusion Detection Scheme (IDS) is being proposed. This system can monitor and analyze the incoming network traffic whether is normal. UNSW-NB 15 dataset is used to validate the machine learning model which is powered by boosting algorithm.Three of the boosting algorithms such as Adaptive Boosting (AdaBoost), Extreme Gradient Boosting (XGBoost) and Gradient Boosting Classifier (GBC) are trained over the six baseline models such as Support Vector Machine (SVM), Decision Tree (DT), K-Nearest Neighborhood (KNN), two variants of Random Forest Model, Gaussian Naive Bayes (GNB). The performance of the IDS is measured in terms of correct analysis of the network traffic as normal or abnormal and the time taken to detect it. As per the observed results the proposed IDS system is providing the best results for XGB model which gives 95.57 % of accuracy and the time taken to do it is come out as 3.03 seconds. The entire experiment is executed both in Central Processing Unit (CPU) and Graphical Processing UnitGPU) environment and a comparative analysis is done in terms of execution time.