Deep Learning Model for Cybersecurity Attack Detection in Cyber-Physical Systems
Mujaheed Abdullahi, Hitham Alhussian, Norshakirah Aziz, Said Jadid Abdulkadir, Yahia Baashar
Abstract
In recent years, there has been an increasing demand for computing devices in cyber-physical systems (CPS), which include smart manufacturing, air intelligent transportation, critical infrastructure, robotic services, and Internet of Things (IoT) infrastructure. Field devices, on the other hand, such as sensors and actuators, which are frequently used for real-time monitoring and prediction, send a large amount of data through the network and communication layers. The CPS is vulnerable to major cybersecurity attacks. To overcome this, there's a need for new deep learning (DL) techniques that can investigate, detect, and respond to changes in such attacks. In this paper, we proposed a DL model for cyber security attack detection in the CPS based on long-short term memory (LSTM). Moreover, the model has been evaluated using real-world datasets from Industrial Control System (ICS) datasets of gas pipelines, which consist of seven attack types with 19 features. The results of the experiment show that the proposed model achieved an accuracy of 98.22% after validation. The paper also presents a recommendation for potential future investigation.