© MAR 2022 | IRE Journals | Volume 5 Issue 9 | ISSN: 2456-8880 IRE 1703291 ICONIC RESEARCH AND ENGINEERING JOURNALS 374 Improving Data Protection in Industrial Control System Networks Using Machine Learning Technique JUDE CHINEDU AKAMADU 1 , PROF. JAMES EKE 2 , EMETU CHUKWUMA KALU 3 1, 2, 3 Department of Electrical and Electronic Engineering Faculty of Engineering Enugu State University of Science and Technology (ESUT) Enugu, Nigeria Abstract- This work was embarked on to combat the security vulnerabilities trending on the present-day Industrial Control Systems (ICS) Supervisory Control and Data Acquisition (SCADA) system infrastructure. This was observed after series of literatures were discussed and research gap was established. The study proposed to solve the problem using machine learning. Data of the ICS Denial of Service Attack was collected and then develop a neural network-based algorithm with it to detect threat on ICS. The training was done using back propagation algorithm. The system was implemented using Mathlab and neural network toolbox. The model was simulated and the result showed good threat detection performance with regression value of 0.973 and detection response time of 12ms which is very good. The percentage improvement when compared with the characterized test bed is 13.3% which is very good. Indexed Terms- Data protection, control system networks, and machine learning I. INTRODUCTION In today’s competitive market domain, companies have demand to improve their process efficiency using a low economic cost industrial automation to optimize productivity. Here the collaborative integrated system performs a big role with the functions of self- organization, rapid deployment, flexibility, and inherent intelligent-processing (Bailey et al., 2013). In the industrial control system (ICS), the operators can perform remote control of various plants such as the continuous stir tank reactor, water treatment plant, power plant, nuclear reactor, biological reactor among others. This remote monitoring and control process is achieved using communication protocols between the human interface machine, sensors, actuators, controllers and Supervisory Control And Data Acquisition (SCADA) system. Historically this SCADA components have dedicated and private networks. Recently the increase in the size and complexity of the modern day industrial facilities has resulted to wide range of remote management, monitoring and supervision of these equipment through an open network called the internet. This exposes the SCADA systems to various forms of network threats and internet attack (Rinaldi et al., 2011). Industrial control system securities have been handled traditionally using the conventional information technology (IT) security practices such as the use of network authentication codes and low level encryption formats. However, the goal of IT security techniques cannot cope with the high monitoring, supervision and physical complex industrial components that make up the SCADA network of an industry. This is to say that in IT while an electronic mailing system can afford short delay without any effect. for an industrial control system, a slight delay can result to devastating effect and environmental hazards, financial losses and even loss of life (Erickson et al., 2019). Machine learning and artificial intelligence techniques have been employed widely to combat these challenges of industrial control security, thus developing intelligent intrusion detection systems using various techniques which will be discussed in the literature review. However, the system accuracy, rate of threat detection, speed of threat detection, system reliability and future threat prediction response abilities have all been called into question recently, due to the dynamic nature of the modern form of industrial threat. Therefore, this research will employ a hybrid machine learning techniques, which will be trained using a