XAI for Securing Cyber Physical Systems

Machine Learning (ML) is heavily leveraged in various fields, specially in Cybersecurity to detect attacks and anomalies. Although these models are capable of achieving high prediction accuracies, they are not interpretable. In fields like medical sciences and cybersecurity, it is essential for the model's predictions to be understandable so that informed decisions can be taken accordingly. A Cyber Physical System (CPS) is controlled or monitored by computer-based algorithms. A CPS is divided into three layers, application control layer, data transmission layer and perception execution layer. Attacks can be performed in any of these layers. Network attacks, cryptographic attacks, physical threats and malicious software injection are some of the possible attacks. In the present work, we used a dataset containing Network Attacks and Physical Threats. The aim was to detect these attacks using ML models and interpret the decisions using Explainable Artificial Intelligence (XAI). By extracting explanations from the model, XAI helps build trust on the model outcome.