Machine Learning-Based Ensemble Network Security System

A vital issue related to security for systems that are interconnected is the undesirable intrusions by agents. Various types of attacks are possible by the intruders. The attacks are associated with network security since the entry to the system can be accomplished through a network. Hence, in a typical organization, attacks can be external or internal attacks. In this work, the possibility of using Machine Learning (ML) techniques to detect network intrusions in combination with other techniques is explored. The outcome of this research will be the automatic application of firewall rules according to the present state of the network to prevent intrusions. Various classifiers are explored and then the Decision Tree (DT) classifier is implemented that functions as a signature-based system. For classification, the network traffic and logs of system in a UNIX based machine are captured. The frequent connections are identified, so that any anomaly in resource usage can be detected. This along with the classifier functions as an ensemble system to prevent network intrusions.