Research On Network Traffic Detection Method For Power Internet Of Things

With the popularization of ubiquitous power Internet of things and the continuous development of computer network,power Internet of things has entered thousands of households.At the same time,the growing cyber attacks also target the huge power Internet of things network.In recent years,the network activities are increasingly rampant,which has caused a great impact on the network security.Malicious traffic detection as an effective defense against network attacks is receiving more and more attention.Traditional machine learning methods are no longer enough to deal with increasingly complex and efficient network attack methods,and they need to be trained with manually extracted features,so their application value is limited.In recent years rapid development in the field of image and text deep learning network provides a new train of thought for the development of network security field,based on partial correlation and time continuity of the network traffic can be used as a convolution of the neural network to identify local features as well as the circulation of the neural network input sequence characteristics,based on this,this article embarks from the power of things network traffic,from the network traffic input to the output to the endto-end strategy,realize the classification of network traffic.The main research contents of this paper are as follows:According to the local correlation of network traffic firstly design convolution neural network adaptive network traffic,by putting the raw binary data into a matrix of network traffic data,as the convolution of the neural network input data,and then after two layers of convolution and pooling operation input connection in the network,finally realizes the output of network traffic prediction results,and in CICIDS2017 training and validation data sets.Experiments show that this method can meet the practical requirements in both accuracy and error analysis.Then according to the power of the Internet of things time continuity length cycle when the neural network based design adaptive network traffic circulation neural network,this method will handle good network stream as a time series as input,one dimensional matrix through memory cells after input entire length to connect to the Internet,in the last layer adopts softmax output forecast probability of each category,and the experimental results show this method can also be used in the accuracy and error analysis for the classification of network traffic,to meet the practical requirements.Finally,based on the above training model,the Tensor Flow reasoning server serves as the classification prediction service,and the client USES Qt as the framework to write the malicious traffic detection system for the power Internet of Things,and realizes the offline local file network traffic classification and online network traffic classification.