Identifying False Data Injection Attacks in Industrial Control Systems using Artificial Neural Networks

Cyber-attacks on Industrial Control Systems (ICS) are growing in recent years. Existing IT-security technologies are not sufficient enough to protect the ICS from the novel attacks. Among several attack types on industrial networks, False Data Injection Attacks (FDIA) are considered as an important class of cyber-attacks against ICS. FDIA injects forged measurement into the control system in hope of misguiding the control algorithm. This abnormal behavior of a single control device or a sensor value in a plant can lead to a huge loss to the company or a disaster in plant environment. Hence, a prior identification of these injected attacks is very important. In this paper, we give an overview about different possible cyber-attacks on ICS followed by the importance and challenges in identifying FDIA w.r.t other attack types. A simulated ICS use case is developed to generate the sensor and actuator signals. An attack injection tool is developed and used to simulate the attacks on to the ICS network. The generated data with injected attacks is used to train and test the performance of the Artificial Neural Networks (ANN) for identifying FDIA. The evaluation of performance parameters shows promising detection accuracies.