Traditional machine learning methods like Support Vector Machines, Random Forest, and gradient boosting have shown strong performance in classifying device behaviors and detecting botnet activity.