This research presents a comparative experimental study of machine learning models for botnet attack detection in Internet of Things network using N-BaIoT dataset. The dataset consists of benign traffic and malicious traffic generated by Mirai and BASHLITE botnets family. A total of 115 traffic features are used for classification. The study compares three representative models of learning, namely Random Forest as a classical machine learning classifier, XGBoost as an advanced model of ensemble learning and fully connected Artificial Neural Network as a deep learning classifier. The evaluation of the models was done looking for performance metrics like accuracy, precision, recall, F1-score etc. As per the experiments conducted, the developed system achieved the best which is XGBoost which produced an accuracy rate of 99.12%, a precision of 98.98%, a 99.26% recall, 99.12% F1-score, a 0.88% false alarm rate, and an AUC which is 0.998. The model also achieved a remarkable inference time of 0.018 ms per sample, outperforming Random Forest with 0.021 ms and ANN at 0.035 ms. These results indicate that XGBoost provides the most effective balance between detection accuracy and computational efficiency, making it a suitable candidate for near-real-time IoT botnet detection at gateway or edge-monitoring levels.
Comparative Evaluation of Supervised Machine Learning Models for IoT Botnet Detection using Random Forest, XGBoost, and ANN
Botnet detection
Classical machine learning
Ensemble learning
Deep learning
Random forest
XGBoost
Cybersecurity