A Multi-class Classification for Detection of IoT Network Attacks Using Machine Learning Models

The adoption of IoT devices is growing due to their versatility and simplicity. The number of security risks associated with these devices has increased as a result of their increased popularity. Therefore, it is crucial to have a reliable IoT network intrusion monitoring system. In order to identify IoT network attacks using machine learning models, this study suggests a multi-class classification method. It includes contemporary attacks that enabled us to categorize them into more than two classes. In this study, we looked at 9 different IoT network assault variations using the UNSW-NB15 dataset and 6 of these 9 attacks are given higher importance during the classification process. The suggested approach includes data preparation, feature selection, and the creation of synthetic data using the CTGAN methodology. After generating synthetic records, the final dataset is used to train and evaluate the efficacy of various machine-learning designs, such as Random forests, Extra trees, Decision trees, and XG Boost, with XGBoost outperforming them all with an accuracy of 96.71. The findings of this study will help to create a more reliable and effective IoT network assault detection system, which will aid in the prevention of possible security vulnerabilities in IoT networks.