Evaluation of Network Intrusion Detection with Feature Selection using Random Forest and Deep Neural Network

Abstract

Modern era relies heavily on the internet for communication and confidential data exchange. Integrity, validity, and security of data transmitted should not be compromised. Intrusion detection plays a role of paramount importance in secure transmission. However, Network intrusions are evolving. It raises the necessity for a more robust and evolving detection system. The main objective of this paper is to compare and contrast the performance of Random Forest and Deep Neural Networks on the CICIDS-2017 dataset to build a robust Network Intrusion Detection System. Data of DDoS, DoS, and PortScan attacks from CICIDS-2017 are considered for analysis. The data is preprocessed then feature selection algorithms are applied and the best split is selected for classification. The paper compares the performance of Random Forest and Deep Neural Networks. It was observed that DNN performs best on CICIDS-2017 data to classify between different attacks on the network.