On Designing Machine Learning Models for Malicious Network Traffic Classification