Multilabel Classification in IoT NIDS: A Proposed Cross Machine Learning Pipeline

Network Intrusion Detection Systems (NIDS) are among the most dynamic cybersecurity assets in most organizations' cyber infrastructure because they provide a reliable means of protection against cyberattacks. In recent years, the expansion of Industry 5.0 technologies has resulted in more IoT devices being connected to the Internet, increasing attack surfaces and placing additional burden on NIDS. In the research world, ML-based solutions for IoT NIDS are evolving very quickly. However, the availability of high-quality publicly available datasets in this area diminishes the potential of this niche. This work compares three IoT-based datasets (ToN-IoT, X-IIoTID and UNSWNB15) through in-depth analysis of their features and architecture. The goal is to assess their readiness for a proposed cross-machine learning pipeline for the purpose of generalizability of the multi-label classification ML model. Using key ML performance metrics such as F1 score, precision, and recall, our results show that our model performed reasonably well in identifying different attack scenarios on each dataset.