DDoS Detection by Using Machine Learning

Abstract

Distributed Denial of Service attack (DDoS) is the most risky attack in network security. DDoS attacks prevent essential services from operating normally for many online applications. With an increasing number of these attacks, the task of detection and mitigation has become increasingly challenging. Among the numerous methods available for detecting Distributed Denial of Service (DDoS) attacks, machine learning techniques have shown great promise in effectively identifying and preventing such attacks. In this project, the machine learning-based model was proposed to detect DDoS attacks. The proposed model used the DDoS-CICIDS2017 dataset with 79 features, and applied four algorithms: Logistic Regression (LR), Support Vector Machine (SVM) with different kernels, Random Forest (RF), and Gradient Boosting (GB). The results highlight the outstanding performance of the Random Forest model, achieving an exceptional 99.99% accuracy, precision, recall, and F1 Score. Notably, this model demonstrated a perfect precision of 100.00%, underscoring its efficacy in accurately classifying DDoS traffic and solidifying its role as a formidable defense against these cyber threats.