A combination of PSO-based feature selection and tree-based classifiers ensemble for intrusion detection systems

Abstract

Due to the numerous attacks over the Internet, several early detection systems have been developed to prevent the network from huge losses. Data mining, soft computing, and machine learning are employed to classify historical network traffic whether anomaly or normal. This paper presents the experimental result of network anomaly detection using particle swarm optimization (PSO) for attribute selection and the ensemble of tree-based classifiers (C4.5, Random Forest, and CART) for classification task. Proposed detection model shows the promising result with detection accuracy and lower positive rate compared to existing ensemble techniques.