Comparative Analysis of Detecting Cyber Attacks Using Systematic Hyperparameter Exploration Method (SHEM)

Industrial sectors such as manufacturing, smart grids, supply chains, food production, and water treatment plants depend heavily on Information and Communication Technology (ICT). New emerging trends and technologies, such as the Industrial Internet of Things (IIoT) and the Industrial Internet of Services (IIoS), interact with traditional, isolated ICT systems. Various critical and complex industries such as manufacturing, electric grids, pharmaceuticals, and water treatment facilities employ “smart” technologies to increase productivity and efficiency and reduce production costs. Despite its benefits, integrating innovative and legacy ICT creates multiple complex interdependencies that require more safe and secure countermeasures. Consequently, rapid evolution has dramatically affected threat landscapes. Owing to the consequences of these threats, this study explored two powerful deep-learning-based algorithms (MLP and BiLSTM). Moreover, the study also investigated the various configurations of MLP and BiLSTM to assess how these fine-tunings impact the performance of the model. Using a Systematic Hyperparameter Exploration Method (SHEM), this study aims to provide valuable insights into how an appropriate configuration can improve performance and make the model more robust for detecting cyber-attacks in smart industries. The results of this study suggest that selecting an appropriate combination of configurations is critical for maximizing model robustness and performance.