| In recent years,with the continuous development of big data and cloud computing technology,computer network technology has become popular in many fields such as economy,politics,medical treatment,and industry,providing convenience to people’s daily life and production process.However,network security issues have also followed.More and more server equipment and illegal programs have buried major hidden dangers for network security.Our various private information is continuously uploaded to the cloud database by various rogue software and websites.At the same time,some criminals obtain people’s personal information through illegal network attacks.In the face of the explosive growth of Internet data,traditional network security detection technology can no longer effectively cope with the current situation,so using an intrusion detection system to prevent network intrusion has become an urgent need.Therefore,this paper analyzes and evaluates the detection performance of various machine learning algorithms in network intrusion detection.By studying the advantages and defects of algorithms in network intrusion detection,a new classification algorithm model is proposed to improve the detection performance of network intrusion detection.The main work of this paper is as follows:(1)The source and characteristics of intrusion data are analyzed,and an intrusion detection model based on machine learning is constructed.The traditional supervised machine learning algorithm was studied,the algorithm was trained with the NSL-KDD dataset and the UNSW-NB15 dataset,and appropriate parameters were selected for different algorithms to learn,and the multi-classification task of network intrusion detection was completed.(2)This paper proposes a clustering-based Borderline-SMOTE algorithm for the imbalance of the intrusion detection data set.Before enhancing the minority class samples,the method first clusters the samples,and then performs clustering on the samples in each cluster.enhanced.This can ensure that only those relatively few minority class samples are selected for operation during enhancement,without affecting those minority class samples that are already sufficient.To a certain extent,it is more conducive to improving the classification effect of minority samples.(3)This paper studies and proposes a new type of intrusion detection model,which uses the convolutional neural network in deep learning as the base classifier,and first preprocesses the original data,including numericalization and normalization.Secondly,a feature selection method based on RFE method and random forest model is proposed,which can effectively extract features useful for classification and reduce the consumption of computing resources.Finally,the CNN-LSTM model is introduced,which can make full use of temporal information and spatial information,and solve the problem of poor generalization ability of a single detection model.The experimental results show that the method proposed in this paper is significantly better than other algorithms in terms of accuracy and recall,and has a greater improvement compared to traditional machine learning algorithms.Among them,the classification accuracy and F1 score of the model on NSL-KDD both reached more than 99.9%.At the same time,the accuracy rate on the UNSW-NB15 dataset also reaches 88.62%,which has excellent detection performance and brings great advantages to the field of intrusion detection. |
