By Eleanor Haas

Abstract: With the conveniences of modern technology comes the increasing risk of cyberattacks, and industrial control systems in infrastructure are particularly vulnerable. While they provide convenience and efficiency, there is also an increasing risk of disruption to these systems, whether by cyberattack, faulty engineering, or natural disasters. This research focuses specifically on pipeline and transportation infrastructure for oil, and how AI can be used to classify data from industrial control systems as normal or abnormal. Two AI models were designed with H2O Flow, a Generalized Linear Model and a Gradient Boosting Machine, and were trained and validated on a sample dataset. The results suggest that using AI to analyze incoming data is, or will soon be, a reliable method of detecting anomalies affecting a system. This is important for identifying cyberattacks, damages, or accidents quickly in our critical infrastructure systems.