Enhanced Botnet Attack Detection using Machine Learning and Neural Networks

In an era characterized by pervasive interconnectivity, Internet of Things (IoT) devices have revolutionized automation, facilitating seamless data exchange. However, this technological leap has triggered heightened concerns about data security amid proliferating cyber-attacks. Given the omnipresence of IoT systems, the need for fortified networks to thwart potential intrusions is imperative. This work focuses on crafting an Intrusion Detection System (IDS) for IoT network security based on Machine learning (ML) and Deep Learning (DL) techniques, utilizing the Bot-IoT dataset from the University of New South Wales. Using a small dataset after preprocessing and feature selection, a comparative analysis of multiclass classification models, including Multinomial Logistic Regression, Artificial Neural Networks, Decision Tree Classifier, and Random Forest Classifier, is conducted. The results highlight the Decision Tree Classifier as the optimal model, with an exceptional accuracy of 99.98%, closely followed by the Random Forest Classifier with an accuracy of 99.91%. Automated hyperparameter tuning using GridSearchCV significantly contributes to enhancing model performance. A thorough evaluation, incorporating multi-classification metrics like Accuracy, Precision, Recall, and F1-score, is done to ensure a comprehensive assessment of the models. This research offers valuable insights into fortifying IoT network security, with implications for safeguarding the integrity of interconnected systems.