Evaluating effectiveness of an embedded system endpoint security technology on energy delivery systems

Abstract

The Internet of Things (IoT) is starting to take off in the modern day tech scene, with almost every user device being connected to a huge network with other devices; industrial energy delivery systems are no different. However, IoT in industry (known as IIoT) severely lags behind conventional IoT networks when it comes to cybersecurity, as IIoT endpoint devices generally lack the same level of memory and computation as conventional IoT endpoints. As a result, the same level of security measurements can't be implemented, and IIoT devices often are vulnerable to malicious users attempting to hack the device. The goal of this research is to create a lightweight software system that protects the endpoint devices from hackers, as well as prevent malicious accesses from impacting other parts of the system. This thesis will focus on the command and control unit, where the aim is to develop a neural net classifier to detect anomalies in network traffic. This thesis details the background surrounding IIoT endpoint devices, and the current attempts that have been made at providing a solution, as well as our approach to solving this problem, using both supervised and unsupervised machine learning approaches, and describe the testing environment with which the experiments were conducted. In the end we discuss how the work contributes to the future progress of IIoT security.