Intrusion Detection Method Of Internet Of Things For Sample Imbalance

The Internet of Things is developing rapidly in various fields such as sensors,smart home and industry.However,Io T devices are complex and diverse.When applied to various fields,Io T attacks occur frequently and in large numbers.Ensuring the security of the Internet of Things is a necessary work for the sustainable development of the country.Io T intrusion detection technology can effectively detect attacks,discover threats in advance,and effectively prevent Io T device damage.Today,deep learning related technologies have been effectively applied in Io T intrusion detection.The traditional Io T intrusion detection system has some shortcomings.The quantity of normal sample data collected by the Io T is much larger than the quantity of abnormal sample data.When training the model,the sample data is often unbalanced,resulting in poor model training effect.At the same time,the Io T model has defects such as insufficient feature extraction,which leads to problems such as low model accuracy.To this end,an efficient intrusion detection method is proposed.The research mainly focuses on the following two aspects:(1)Aiming at the problem of imbalance between normal sample and abnormal samples in the Internet of Things,a method of using generative adversarial network to deal with sample imbalance is proposed.In the real Io T environment,the number of normal sample data is much higher than the number of abnormal sample data.In the process of training the model,the lack of abnormal sample data leads to insufficient training model and reduces the training efficiency of the model.This paper uses deep convolutional generative adversarial network to expand abnormal sample data,and generates four types of attack sample data for NSL-KDD data set,so that each type of sample data set reaches a balanced state,thereby improving the efficiency of Io T intrusion detection model.(2)Aiming at the problems of low accuracy and low accuracy of Io T intrusion detection,an intrusion detection model is designed.First,the preprocessed data is passed through DCGAN to make the sample reach a balanced state,Then use the GRU to learn the time series features of the sample data,extract the real sample and generate the features of abnormal samples,The extracted features are input into the Res Net,and the Res Net is used to detect the sample data.When the Res Net is trained in a model with deeper layers,the problem of degradation will not occur,and the accuracy of the model is improved.The simulation test results show that the accuracy of the proposed intrusion detection method reaches 94.56%,And compared with LSTM-ResNet?GRU?LSTM and SVM to verify the effectiveness of the model.