Enhancing IoT Attack Detection Through Ensemble-Based Multiclass Attacks Classification

Interconnected nodes communicate, exchange, and transfer data through multiple network protocols to form the Internet of Things (IoT). Study shows that these protocols' ease of exploitation makes them vulnerable to cyberattacks on data transmission. In this study, we anticipate improving IDS detection efficiency and adding to the literature-multiclass classification of IoT traffic attacks to enhance IDS efficiency. Multi-class supervised ML methods and ensemble classifiers are used. TON-IoT network traffic was used for model training. Five trained ML-Multiclass supervised models-Random Forest, Decision Tree, XGBoost, Extra Tree, and K-Nearest Neighbor-are the most accurate. Voting and stacking ensembles use these five classifiers. The evaluation measures were used to compare the ensemble techniques' classification effectiveness. Ensemble classifiers outperformed multiclass models. Ensemble learning systems that use varied learning processes with different capacities improved this. Combining these methodologies improved classification reliability and reduced classification mistakes. Experimental data indicates that the proposed framework can enhance Intrusion Detection System efficiency with 0.98923 accuracy.