ELA-NIDS: An Improved Ensemble Learning Network Intrusion Detection System for IIoT
Kumar Saurabh, Uphar Singh, V Sai Vamshi, Ranjana Vyas, Om Prakash Vyas · 2024
The Industrial Internet of Things (IIoT) has revolutionised the industrial sector by establishing connectivity between machines and devices through the internet. The heightened level of connectivity has concomitantly amplified the vulnerability to cyber-attacks, which pose a threat to the uninterrupted functioning of organisations leading to economic and reputational damages for businesses, as well as the misappropriation of confidential data. Machine Learning(ML) based Network Intrusion Detection Systems(NIDS) have gained notable popularity to safeguard against security breaches by monitoring anomalous behaviours. The feature sets of the presently accessible NIDS datasets exhibit significant differences which significantly influences the performance of ML models making them unreliable. The proposed approach is based on the NetFlow-based (NF-ToN-IoT-v2) dataset, which is a standardised feature set for NIDS datasets. The primary contribution of this study lies in its capacity to systematically and dependably assess the effectiveness of Ensemble and ML-based traffic classifiers across a diverse array of network topologies, attacks, and other relevant factors. The experimental findings indicate that XG-Boost exhibited a maximum accuracy of 96.01 % in detecting various forms of attacks in an IIoT network.