Cybersecurity threats have been rapidly increasing, specifically in the chemical industry that contains critical, high-value infrastructure vulnerable to attack in regions. Cyber attacks that can result in the loss of control of chemical processes can physically cause harm to the real world. Thus, the need for a robust system in operational technology (OT) to counter the risks posed by cyber attackers is urgent. Cyber attacks can be categorized into both sensor and actuator cyber attacks. Sensor attacks are designed to tamper with the data collected by sensors in a system, such as altering temperature readings. Conversely, actuator cyber attacks mediate directional changes to operations of a system, such as valves, motors, or pumps, without being detected by sensors. Detecting both sensor and actuator attacks involves monitoring system data to identify anomalous behavior. To improve the distinction between normal and anomalous behavior, a neural network (NN) model is created to classify three types of closed loop systems: nominal closed-loop systems, faulty closed-loop systems, and closed-loop systems under cyber-attack. Developing a NN model to be implemented in cyber attack detection systems streamlines cybersecurity and predictive control.
