A Comparative Analysis of Black-Box and White-Box Models for IOT Botnet Detection

Botnets pose a significant threat to network security, especially to the potential server crashes they can cause. The proposed study focuses on methods for identifying the presence of and the type of botnet activity, by implementing machine learning techniques, and consecutively comparing the black-box and white-box approaches employed. Utilizing the CTU-13 and N-BaIoT datasets, the study involves data preprocessing and a comparative analysis of various models. The LGBM Classifier emerges as the top-performing model for detecting the presence of a botnet, while the XGBoost Classifier excels in identifying the specific type of botnet attack based on different performance evaluation metrics.