Research And Implementation Of Instrusion Detection Based On Deep Learning

With the popularity of the Internet and mobile Internet,more and more people share information through the Internet.The sharing of network information brings convenience to people's life,and it also poses a huge threat to the security of user information.Intrusion detection technology is an important part of the network information security defense system.Unlike passive detection systems such as firewalls,it can actively detect potential intrusion information in the network and provide more comprehensive information protection for users.Traditional intrusion detection methods learn by matching the features of the feature library or using the classification and clustering methods.The above detection methods rely heavily on the selection of features.Therefore,in order to make intrusion detection have strong adaptability and high accuracy,it is necessary to have a deeper understanding of features.At the same time,there is a problem of data imbalance in Intrusion detection,which makes the effect of intrusion detection have a large bias,and the detection effect on the category with fewer intrusive viruses is very poor.Moreover,in the early stage of the emergence of invasive viruses,there are very few invasive data and even no data can be trained,which makes the detection of invasive viruses face great challenges.In order to respond to these above challenges in intrusion detection,this paper uses deep learning methods to detect intrusion data comprehensively.The main contributions of this paper are as follows:This paper proposes to use intrusion detection algorithm based on stacked sparse autoencoder network to solve the problem of massive intrusion data detection.The stacked sparse autoencoder network is used to reduce the dimensionality of a large number of high-dimensional,unlabeled original data,so as to obtain the deep feature representation of the original data.Compared with multilayer perceptron network,the F1 value has been significantly improved.According the problem of intrusion data class imbalance,at the data level,this paper adopts a staged sampling algorithm to sample unbalanced data.At the algorithm level,an ensemble learning algorithm based on stacked sparse autoencoder network is proposed,which improves the detection effect of class imbalance intrusion data.At the same time,a weighted fusion method of multi-classification ensemble learning is proposed to improve the intrusion detection effect of ensemble learning,which makes the F1 value of the categories with less data increase obviously.In order to solve the problem of absolute imbalance between intrusion data categories,there are very few training samples in the early stage of the emergence of invasive virus.This paper proposes one-class learning method based on stacked sparse autoencoder reconstruction error,which can detect the invasive virus well in the early stage.Compared with the ensemble learning method,the F1 value of the categories with very little data increased by 27.4%.