Intrusive Detection in Nuclear Systems using Supervised and Unsupervised M1 Methodologies

In order to safeguard vital infrastructure, nuclear systems must employ intrusion detection. In such a dynamic environment, it may be challenging for conventional intrusion detection methods to identify modern, adaptive attacks. Utilizing machine learning to enhance nuclear intrusion detection is an emergent field of study. This study provides an overview of machine learning methods, including examples of their application, for nuclear facility intrusion detection. Both unsupervised learning strategy anomaly detection (Isolation Forest), and supervised learning algorithm Convolutional Neural Networks. This conducts a series of case studies using data from operational nuclear systems to evaluate the efficacy of these machine learning methods. The present experimental data, evaluate the efficacy of different algorithms, and discuss to select and preprocess features. The study also discusses the challenges and potential solutions associated with the use of machine learning for the detection of nuclear system intrusions. Both models perform well in detecting intrusions, with CNN outperforming the baseline and Isolation Forest in terms of performance metrics. This demonstrates how beneficial CNN is for enhancing the intrusion detection of nuclear systems by identifying subtle patterns and features in vast quantities of data. Our findings shed light on the potential of machine learning to improve the security of nuclear systems and provide valuable lessons for future cyber security professionals.