Research On DDoS Attack Detection Under Hybrid Model Of Autoencoders And RBM

With the continuous development and improvement of cloud computing technology and Internet technology,network security has gradually risen to the focus of attention.In the rapid change of information age,the methods of network attack have been varied,and its own attack scale,rapid change,complexity and variety of problems have risen sharply.Distributed denial-of-service(DDoS)attacks are designed to prevent legitimate users from accessing desired target resources.How to effectively prevent such DDoS attacks and how to solve the security-related problems in cloud computing has become the difficulty and challenge point that people want to solve under the background of such a rapid development of information age.Aiming at the above problems,this paper proposes a study on DDoS attack detection under a hybrid model of Autoencoders and restricted Boltzmann machine.First,the main network features are extracted from a large number of network traffic data,and then the data are processed to provide a good data source for the later design module.Secondly,the main feature selection is carried out after the data preprocessing is completed,and the feature selection for different data sets is different.and then,according to the data sets of different dimensions,the automatic encoder layer model with higher dimension and high accuracy is selected,and the combination collocation is carried out by setting a restricted Boltzmann machine(RBM)between each two layers.By the coding layer of auto encoder(Autoencoders)to achieve the effect of dimension reduction,every two layers with restricted Boltzmann machine can adjust the model well.The advantage of selecting RBM is that it is a model based on energy function.After obtaining the energy function,a series of related probability distribution models can be introduced by using the energy function.Finally,the parameters are modified on the basis of the existing model,so that the data set can achieve the desired output.Finally,the feasibility and credibility of the system are verified by model and experiment.The experimental results show that the method greatly reduces the learning time and increases the accuracy by selecting and processing the features of different data sets.The practicability of the model is obtained by comparing the data values of the existing data sets.So the research of DDoS attack detection under the hybrid model based on Autoencoder and RBM is still very good to improve the prediction accuracy of DDoS attack detection,which has the advantages of good applicability and high accuracy compared with the traditional intrusion detection methods.