IoT Security Viewer System Using Machine Learning

Abstract

Recently, Internet of Things (IoT) are spreading and various things are connected to the Internet. As the results, it is possible to get various data and operate the devices remotely. With spreading of IoT, attacks of malwares like Mirai and Hajime which target the IoT devices are increasing. For instance, a large-scale DDoS attack by infected devices as a springboard has occurred. Network monitoring tools for IoT devices have also been developed as attack measures against IoT devices. However, the developed tools have many types of software to be introduced and visualization of the network topology are not performed, so there is a problem that visually instantaneous abnormalities cannot be recognized. In this research, we develop a system that detects abnormality by machine learning, visualize a network topology, and notifies abnormality by alert visualization on the network topology. We measure the accuracy of abnormality detection and the real time property of visualization of alerts by actual machine experiments and show the effectiveness of the proposed system.