A Network Intrusion Detection Model based on the Combination of ReliefF and Borderline-SMOTE

Abstract

To ensure the security of network environment, intrusion detection technology should be adopted to effectively resist network attacks. To solve the problem of imbalanced data distribution existing in the intrusion detection data set, this paper proposes a novel two-level network intrusion detection model based on the combination of ReliefF algorithm and borderline-SMOTE oversampling technique. In the proposed model, ReliefF algorithm is first used to select features that can better express the imbalanced data distribution, and then the borderline-SMOTE oversampling technique is used to oversample the misclassified minority class samples. The proposed two-level intrusion detection model contains two base classifiers. In this paper, three different types of base classifiers, KNN, C4.5 and NB, are combined in pairs and tested by 10-fold cross validation. NSL-KDD data set was used to verify the model. Experimental results show that the system can perform well on imbalanced network intrusion detection data set and significantly improve the detection accuracy of minority samples.