An Intrusion Detection Model for CICIDS-2017 Dataset Using Machine Learning Algorithms
Shailesh Singh Panwar, Y. P. Raiwani, Lokesh Singh Panwar
Abstract
Due to excessive use of the internet, keeping the network secure and transferring data securely across networks has become a difficult task. To recognize distinct types of network (internet) attacks, need Machine learning strategies, to develop various types of the intrusion detection system. The main reason behind using different types of intrusion detection strategies is that the attackers constantly receiving information about the network traffic. Therefore, the intrusion detection system is used to easily identify the attacks and to thwart those attacks. In this paper, we have used the CICIDS-2017 dataset, which is a labeled dataset, for analyzing the result of eight different supervised classification techniques (GaussianNB (GNB), BernoulliNB (BNB), Decision Tree, KNN, Logistic Regression, SVM, Random Forest and SGD). The dataset is divided into five different days; each has different types of class attacks (DoS, PortScan, Botnet, Web Attacks, Infiltration, Heartbleed and DDoS). We have proposed a model with three stages. The first stage data-preprocessing, which includes several steps such as extracting independent and dependent variable, splitting dataset, finding missing value, feature scaling on the dataset and encoding categorical data, then in feature selection stage, finding out unique features using RFE (Recursive feature elimination) technique. Finally, test the result of eight different supervised classification techniques using cross-validation on the CICIDS-2017 dataset.