Detection of Botnet Attack in IoT Environment

This study presents a comprehensive framework for botnet detection that leverages advanced machine-learning techniques to enhance accuracy and robustness. The framework integrates bagging methods, such as Random Forest and Bagged Decision Trees, alongside boosting algorithms like XGBoost, LightGBM, to achieve superior model generalization. To optimize feature selection, Recursive Feature Elimination (RFE) and Principal Component Analysis (PCA) are employed, ensuring the extraction of the most relevant features from network traffic data. The system addresses data quality challenges through rigorous cleaning, normalization, and class imbalance correction using the Synthetic Minority Over-sampling Technique (SMOTE). Evaluated on the UNSW-NB15 dataset, the proposed framework demonstrates exceptional performance, achieving high accuracy and precision-recall area. The results highlight its effectiveness in detecting botnet attacks and its potential as a scalable solution for enhancing network security.