IoT Botnet Attacks Detection and Classification Based on Ensemble Learning

With the vigorous development of the IoT, botnet attacks against the IoT have become more frequent and diverse, and the research on attack prevention and detection has become more difficult. This paper proposes an IoT botnet attack detection model based on feature selection and ensemble learning. Specifically, the model first reduces the feature dimension of the data by selecting appropriate feature sets. Then, an optimized LightGBM classifier and Naive Bayes classifier are integrated to improve the robustness of the model and accuracy. Finally, a meta classifier TPE-LightGBM is used to combine the result of the classifiers and make the final classification. The performance of the proposed model is tested on the N-BaIoT dataset. The experimental results show that it has a detection rate of 99.97% and shortens the training time by 63.75% at most, improves 4.95% to the conventional methods.