Network Intrusion Detection Using Distributed Genetic Random Forest Method

Abstract

The increasing use of internet networks has led to an increase in threats and new attacks every day. To detect an anomaly or misused detection, an Intrusion Detection System (IDS) has been proposed as an important component of a secure network. Because of the model-free properties that enable them to identify the network pattern and discover whether they are normal or malicious, the Machine-learning technique has been useful in the area of intrusion detection. Different types of machine learning models have been exploited in anomaly-based IDS. There is a growing demand for reliable and real-life attack data sets within the research community.To improve the detection rate of NIDSs and reduce the false alarm rate, a lot of works apply a variety of methods of machine learning on NIDS. The Proposed Distributed Genetic Random Forest Method (DGRFM) to build intrusion detection systems exhibits a superior performance than the traditional genetic methods. The proposed multiple-level hybrid classifier to build NIDS to enhance RF with unlabeled data and trained to detect previously “unseen” attacks. The huge amount of network data with the unbalanced distribution of normal and anomaly behaviors lead to the problems of low detection rate and high false alarm rate in most NIDS.