Hybrid Architecture for Distributed Intrusion Detection System Using Semi-supervised Classifiers in Ensemble Approach
Shraddha R. Khonde, Venugopal Ulagamuthalvi · Advances in Modelling and Analysis B · 2020
Security of data is becoming a big treat today because of modern attacks. All the data passing through network is at risk as intruders can easily access and modify data. Security to the network is provided using Intrusion Detection System (IDS) which helps to monitor and analyze each packet entering or passing through the network. In this paper hybrid architecture for IDS is proposed which can work as an intelligent system in distributed environment. Proposed system makes use of semi-supervised machine learning classifiers into an ensemble approach. Classifiers used are Support vector machine, decision tree and k-nearest neighbor. Ensemble of this classifier is done and final prediction is given by majority voting algorithm. This system makes use of feature selection technique to reduce number of features used for training various classifiers. Experiments are conducted on NSL-KDD dataset. From results it is observed that ensemble technique increases accuracy by 3% and reduces false alarm rate by 0.05. System performance improves if used in ensemble approach as compare to individual classifier.