cybclass: classification approach for cybersecurity in industry 4.0

Abstract

Cyberinfrastructure is characterized by a large amount of emerging and dynamic information, requiring a large number of cyber-criminals trying to acquire information, data mining, machine learning, measurements, and other interdisciplinary skills to meet the cybersecurity issues in Industry 4.0. Machine learning and information mining play an important role in cybersecurity, and unstable information frequently has a high-dimensional feature space. The presence of several noisy characteristics among high-dimensional features might impede and degrade classifier performance. To address this issue, feature selection and subspace methods have been put out and assessed during the past few years. In this paper, four classification techniques and a feature selection strategy are implemented to detect attacks that threaten Industry 4.0. These techniques are Random Forest (RF), Decision Trees (J48), Support Vector Machines (SVM), and Naive Bayes (NB) with Feature Selection Strategy (CFS). Several experiments have been performed using the train and test NSL-KDD datasets with good results. These are based on four categories: Denial of Service (DoS) attack, Probing Attack, User-to-Root (U2R) attack, and Remote-to-Local (R2L) attack. To improve the detection rate of these attacks, a strategy combining multiple classification algorithms is implemented.