Abstract
With the rapid growth of technology, the Internet's use and the number of devices connected to it are growing at a breakneck pace. As a result of this development, network traffic has increased in volume and has become more vulnerable. The focus has been on the development of learning intrusion detection systems in order to detect sophisticated and undetected threats. Because machine learning-based models achieve great accuracy in a short amount of time, they are commonly utilized in intrusion detection systems. Multiple classifications were made in this study to detect assaults on network traffic using machine learning. The model was created using the CICIDS2017 data set, which comprises both current and historical attacks. The high-performance computer was used to rapidly conduct tests on the CICIDS2017 data set, which contains around 2.8 million rows of data. We improved the performance of the machine learning models we developed by cleaning, normalizing, oversampling for an unbalanced number of labels, and reducing the size of the data set using feature selection methods. The random forest, decision tree, logistic regression, and Naive Bayes classifiers were all implemented on the pre-processed data set, and it was observed that the random forest classifier had the highest accuracy of 99.94%.