What are the primary challenges in using machine learning for DDoS detection?

Several challenges exist. Data imbalance (legitimate traffic vastly outnumbering malicious traffic) is a major hurdle. Feature engineering to effectively capture DDoS attack characteristics is complex. The dynamic nature of DDoS attacks requires models that can adapt to evolving attack patterns, necessitating online learning or frequent retraining. High computational cost of complex models can be a limiting factor for real-time detection. Furthermore, adversarial attacks can be used to evade machine learning models.