HIERARCHICAL ML MODEL FOR DISTRIBUTED DDOS ATTACK CLASSIFICATION AND HYPERPARAMETER TUNING

Abstract

The security of networks is seriously threatened by sophisticated DDoS attacks. In order to increase the precision of distant detection and classification, this article describes a hierarchical machine learning methodology. The system starts with a core layer that detects fundamental issues, then expands classifications into multiple DDoS assault categories. By eliminating extraneous data and increasing efficiency, the approach uses mutual information and correlationbased filters to identify the most relevant qualities. Bayesian optimization for hyperparameters enables improved detection compared to conventional techniques like grid and random search. Experiments using datasets like CIC-DDoS2019 show that this approach lowers false positives while significantly improving accuracy, recall, and F1-score. This cutting-edge and adaptable technology works in real time and can defend modern network systems against dynamic DDoS attacks.