Feature Selection based Performance Analysis of Machine Learning Algorithms in Network Intrusion Detection

Abstract

Information and data security is one of the most challenging tasks for the massive-scale digital revolution all over the world. The very first step to secure our data is to identify invasive conduct and intrusive behavior. However, due to the high scalability of most modern systems and the complex nature of the attacks, the traditional detection system is less reliable. To overcome this challenge, it is necessary to build intelligent and adaptive intrusion detection technologies. That is why this research developed an intrusion detection system using commonly used ML algorithms and analyzed performance from different perspectives. In our pipeline, this exploration applied both supervised and unsupervised learning algorithms. The training and test data were split in multiple ways to evaluate the performance of the models. From the experimental results, it was found that the Light Gradient Boosting Machine (LightGBM) performs better in our context in terms of both precision and recall.