High-Accurate Application-Layer DDoS Attack Detection Using Machine Learning

Distributed Denial-of-Service (DDoS) attacks continue to escalate in frequency and severity, disrupting system access and network functionality. This paper addresses the evolving landscape of DDoS attacks, particularly the shift towards sophisticated application layer DDoS attacks, which challenge traditional detection methods. A novel machine learning approach is proposed using a combination of random forest, decision tree, and genetic algorithm, along with a multilayer perceptron classifier, to detect five distinct categories of network traffic, including four variants of DDoS attacks and benign traffic. The methodology involves meticulous data preprocessing, optimal feature selection through RF and GADT, min-max scaling, and multi-layer Perceptron classifier classification. The effectiveness of the proposed approach, which is demonstrated using the CICIDS2017 dataset, achieves exceptional results, with accuracy reaching 99.1%, precision of up to 92.2%, F1 score of up to 95%, and recall of up to 98.4%. The obtained findings highlight the potential for accurate AppDDoS attack detection using a streamlined feature subset, contributing to enhanced network security.