A machine learning based approach for DDos attack detection

Abstract

The modern era saw the rise of technologies in almost every sector. Computers are gradually becoming faster and smaller, also allowing people to utilize them almost everywhere. As computer technology has become an important part to simplify our life at work, the security of computer networks is one of the hardest challenges for the technology experts to overcome. Network security is a must because it protects private information from online attacks and upholds the dependability of the network. In this study, after reviewing a few previous papers and research works, we decided to work on the detection process of DDoS that can be used on the web or server security. Working on the datasets (CICDDoS2019) to merge them and create a new taxonomy for detecting DDoS attacks was our primary step. Then, the data were generated for the two types of attack which are Reflection based and Exploitation based to reduce the time consumption. Thirdly, using the generated dataset, some Machine Learning based models and classifiers have been implemented on important features that have the most contributions. For getting a better accuracy rate, Random Forest, Naive Bayes, Decision Tree and XGBoost model were applied. Finally, we get a better accuracy rate with these models to detect the attack in a reduced amount of time.