Enhancing Industrial IoT Intrusion Detection with Hyperparameter Optimization
Pranav Sahu, Om Prakash Vyas, Rishita Barnwal, Ayushi Singla, Priyanshu priyanshu · 2024
The Industrial Internet of Things (IIoT) has changed the way of industrial processes, bringing a new era of connectivity and automation. But as these systems are becoming more interconnected, they are also becoming bigger targets for cyberattacks which is why Intrusion Detection Systems (IDS) are so important, to spot and react to these threats. However, for IDS to be truly effective in a useful sense, they need to keep pace with the constantly changing nature of cyber attackers. In this study, we explore how hyperparameter optimization can improve IDS performance in IIoT settings. Our study finetunes various parameters to enhance IDS detection of both known and unknown (zero-day) attacks. We tested a variety of hyperparameter optimization methods to see which ones could boost the accuracy, efficiency, and the robustness of the IDS. For the known attacks, we focused on improving classification accuracy, while for unknown attacks, we used k-means clustering to find unusual patterns which may indicate a threat. By fine-tuning these parameters, we aim to create IDS that is more resilient and adaptable to the ever-evolving landscape of cyber threats. We compared our approach to existing methods to evaluate its effectiveness in terms of detection rates and false positives and our results show that hyperparameter optimization can make a clear significant difference in performance of our IDS. This finding could be a game-changer for security in IIoT environments, offering a more robust way to guard against cyberattacks. The insights from this study could also guide future research on optimizing machine learning-based IDS for industrial applications.