Internet of Things Security Analytics and Solutions with Deep Learning

Abstract

This study presents a new solution applied to defending networks of Internet of Things (IoT) devices. It aims at providing a comprehensive solution to defending the IoT and establishing a protocol for IoT security. Recent attacks that compromised over 120 million devices highlighted the need for enhancing IoT security. This paper introduces the adoption of deep learning for critical security applications by utilizing snapshots of network traffic from nine real-world IoT devices. Furthermore, a set of tools, and in particular, Support Vector Machines (SVM), Random Forest and Deep Neural Network (DNN) algorithms are tested and compared against one another to determine which is the most deployable and provide the highest accuracy of anomaly detection. The obtained results exhibited that all three tested algorithms provided high accuracy. However, the deep neural network provides the highest coefficient of determination compared to the other tested models, making DNN more suitable for this type of applications. Finally, the DNN's learning autonomy feature allows omission of humans from the loop resulting in time efficient real-world algorithm.