A Novel Method to Detect Cyber Attacks in IoT/IIoT Devices on the Modbus Protocol using Deep Learning
Thierno Gueye, Yanen Wang, Mudassar Rehman, Ray Tahir Mushtaq · Research Square · 2022
Abstract The dominant Intrusion Detection Models in the field of IoT/IIoT cybersecurity use network-based datasets. This paper proposes a way to perform Intrusion Detection with just the values in the registers of a device over a Modbus communication protocol. The proposed method of performing Intrusion Detection is by introducing an Embedding Layer into an otherwise simple neural network. Three neural networks were designed for the experiment and trained for both a binary classification of whether an attack occurred or not, and a multi-class classification of different types of attacks. The models performed better than the results previously reported without the use of Embedding Layers. This paper shows that a neural network with an Embedding Layer can effectively be used to model not only whether an attack occurred on a device, but also the class of attack that occurred.