Supporting Cyber-Attacks and System Anomaly Detection Research with an Industry 4.0 Dataset

Abstract

In the era of Industry 4.0, smart devices in the manufacturing industry have become increasingly used and interconnected. However, this also increases the possibility of anomaly and cyber-attacks in the heterogeneous smart systems, resulting in failure which may have a cascading effect on different machines in the factory. Physical damage or business interruption can be caused by cyber-attacks in the manufacturing industry. Machine learning (ML) is an important enabling technology in Industry 4.0 and it can be a solution to the problem aforementioned. But there is a lack of publicly available datasets that include network and Operation technologies (OT) data containing different types of cyber-attacks in the context of Industry 4.0 manufacturing systems. Moreover, many researchers do not release the dataset they utilise which makes it difficult to benchmark or compare the work of different researchers. This paper presents a dataset which is acquired from a contemporary and realistic Industry 4.0 manufacturing system. The dataset comprises of seven different scenarios including normal operation, a range of cyber-attacks and anomalies caused by disgruntled employees as well as errors in manufacturing operations. Using the dataset, we train three machine learning models to detect attacks and anomalies, before presenting the results. We conclude that ML classifiers trained by our dataset perform well in detecting most of the attack types and anomalies. Physical and network data can be a good combination to build a robust system for detecting attacks and anomalies.