Performance Analysis of Machine Learning Techniques in Network Intrusion Detection

Abstract

 A lot of sensitive data is being transmitted over the internet nowadays, which leads to increased
risks of network attacks. To identify suspicious and malicious activities to secure internal networks,
intrusion detection systems aim to recognize unusual access or attacks to the network. Machine learning
technology can play a vital role in a scheme to detect intrusion. It is a technology that is based on
classification and prediction, to deal with security threats. In this work, we focus on significant feature
selection and classification using four machine learning algorithms. Adaptive Boost (AdaBoost), Gradient
Boosting, Random Forest, and Decision Tree classification techniques have been tested on the dataset of
network intrusion detection which is collected from Kaggle. In our analysis, Gradient Boosting outperforms
considering the F1-score. Therefore, this machine learning technique can be utilized to implement an
intelligent intrusion detection system.