Detection of distributed denial of service (DDoS) attacks using machine learning methods

With the digitalizing world, the uninterrupted provision of services provided over the internet is of great importance, especially in systems such as hospitals, banking, energy, etc. There are many attack methods to disrupt or disable these services. Denial of service attacks (DoS), one of these methods, are becoming more and more complex and difficult to detect, but organizing such attacks is becoming very easy and low-cost thanks to many tools. With very little knowledge and skills, attackers can perform distributed denial-of-service (DDoS) attacks on target systems and render them unavailable, sometimes for a short time and sometimes for days. In this thesis, K-Nearest Neighbor, decision trees and support vector machines are used to classify CIC-DDoS2019, the most recent and comprehensive attack dataset available online Decision trees approach was the most successful model with a value of 0.99 in accuracy, precision, recall, f-1 score metrics. Support vector machines showed the worst performance with accuracy, precision, recall, f-1 score values of 0.88, 0.90, 0.89, 0.89 respectively. After these results, the machine learning algorithms used were installed on the EN715 Jetson Nano artificial intelligence kit and run on this kit. Given the parameters and machine learning techniques used, this thesis will be a useful resource for researchers considering using these techniques to categorize DDoS attacks. However, in this thesis, an artificial intelligence kit is used at the initial level to categorize DDoS attacks. Basic information is provided for researchers who will conduct more advanced studies on the artificial intelligence kit.