Research On Intrusion Detection Technology Optimized With A Heuristic Feature Selection Method

With the rapid development of Internet technology,despite the tremendous convenience it brings to human’s lives,the openness of it also leads to the constant threat of external intrusion.How to identify these intrusions accurately and quickly has become a new problem in the field of cyberspace security.In order to deal with the problem of intrusion detection,a large number of studies have introduced machine learning techniques.Network intrusion detection based on machine learning can not only improve the accuracy of intrusion detection,but also greatly improve the efficiency of intrusion detection.However,there are several challenges the machine learning technique based network intrusion detection methods faced.On the one hand,the network data generally has the characteristics of large network data and complex features,which makes it’s hard to shorten the detection time and reduce the false alarm rate in when using machine learning models.On the other hand,network data is extremely unbalanced in nature,which easily leads to the risk of overfitting and ignoring minority data in machine learning models.In this paper,we propose a heuristic feature selection algorithm,namely SSA-CFS,to deal with difficulties which the network intrusion detection method faced.The SSACFS algorithm combines the SSA algorithm and the CFS algorithm for feature selection,and then make use of the algorithm MSMOTE and the two-layer classification model to build a network intrusion detection model.The work and contribution of this paper are summarized as follows:Firstly,we designed a heuristic feature selection algorithm named SSA-CFS.Solving the problems of high dimensionality of network data and many redundant features is the key to improving the classification accuracy and efficiency of network intrusion detection models based on machine learning.Thus,in this paper,we propose SSA-CFS.This algorithm uses the CFS algorithm as the evaluation algorithm and the SSA algorithm as the feature search algorithm.It has the advantages of strong optimization effect and fast convergence speed.The results of our experiments show that compare with the existing heuristic feature selection algorithm,SSA-CFS can effectively improve the efficiency and convergence speed of feature selection.We also apply SSA-CFS to different machine learning models,The results of experiments show that there is a certain improvement in classification accuracy and efficiency compared with existing feature selection algorithms and no feature selection.Secondly,we constructed the network intrusion detection method based on the heuristic feature selection algorithm.Due to the large difference between the number of normal data and abnormal data in network data,data balance in different types of data is an indispensable step in the network intrusion detection model.In this paper,we proposed an imbalanced data processing technology based on Mahalanobis distance and SMOTE algorithm,which could improve the detection rate of minority classes by generating minority class data closer to the boundary.Subsequently,a network intrusion detection model based on heuristic feature selection was designed.Through the comprehensive application of heuristic feature selection algorithm SSA-CFS and an improved imbalanced data processing algorithm MSMOTE,the accuracy of multi-class network intrusion detection was further improved.