Detection and Mitigation of DDoS Attacks in Network Traffic Using Machine Learning Techniques

Distributed Denial of Service (DDoS) attacks accelerate to a critical risk to online services, requiring advanced detection mechanisms. In this paper, a machine learning-based approach is proposed to detect the DDoS attacks using LSTM, SVM, and logistic regression models trained on a dataset of DDoS attacks published by Bennet University. This paper shows that the proposed models achieve high accuracy and low false positive rates, outperforming traditional methods. Furthermore, the accuracy of LSTM, SVM, and logistic regression models are compared and is observed that LSTM achieves the highest accuracy, followed by SVM and logistic regression. As a part of the evaluation process, the best model for the proposed network has been trained and selected which detects and mitigates the network attack. LSTM obtained a powerful veracity score of 99.04%, while SVM and Logistic Regression completed veracity scores of 97.4% and 83.93%, individually. Therefore, LSTM is an active model to label SDN localized DDoS attacks. Additionally, this paper demonstrates the potential of machine learning algorithms in detecting DDoS attacks and offers insights for developing more effective DDoS security systems.