An Adaptive Machine Learning-based Threat Detection Framework for Industrial Communication Networks

The development in sophisticated sensory devices has largely revolutionized industrial control systems (ICS). With the popularity of the fourth industrial revolution, which is also referred to as Industry 4.0, more large-scale industrial communication systems are being interconnected by leveraging information communication technology (ICT) paradigms. These new paradigms have resulted in a sustainable improvement in the conventional production systems by facilitating timely and enhanced production rates. The ICS embeds service oriented architecture (SOA), which comprise of supervisory control and data acquisition (SCADA) systems as its backbone. However, such systems have often fallen prey to cyber-attacks leading to temporary failures or complete physical damage to the ICS. In this perspective, this work suggests the adoption of particle swarm optimization (PSO) method and artificial neural network (ANN) model for identifying cyber-attacks associated with ICS. This work considers the gas pipeline control system data to assess the efficacy of the proposed approach in conjunction with other models. It is observed that the proposed approach provides a prediction accuracy of 98.87 % and respective precision and recall values of 97.80 % and 95.79 %.