Detecting Network Anomalies and Intrusions in Communication Networks

Abstract

Detecting anomalies and intrusions in communication networks is of great interest in cyber security. In this paper, we use Support Vector Machine (SVM) and Broad Learning System (BLS) supervised machine learning approaches to detect anomalies and intrusions in datasets collected from packet data networks. The developed models are trained and tested using data from the Internet routing tables, a simulated air force base network, and an experimental testbed. These datasets contain records of both intrusions and regular traffic data. We compare the two machine learning algorithms based on accuracy, F-Score, and training time.