DDoS Attack Detection using Machine Learning Techniques

Aim of this manuscript is to design and develop a real-time system for detecting DDoS attacks using various existing machine learning algorithms. In this script, We have used CICDDoS2019 dataset because it contains instances of both normal and malicious network traffic, machine learning algorithms like Logistic Regression, K-Nearest Neighbour, Random Forest, Support Vector Machine, and Naive Bayes was trained to distinguish the traffic whether it is benign or attack. In the beginning data was cleaned, normalized and then feature selection was done manually to enhance system's accuracy and efficiency. Then the models were assessed by metrics like accuracy, F1-score, recall and execution time. The results indicated that Random Forest outperformed all other algorithms in detecting attacks, while Decision Tree had the lowest execution time with similar results. While Random Forest can be taken as a reliable model, Decision Tree can also be used for faster classification of attacks.