Abnormal Traffic Detection Method Based on MobileNet-V2 Migration Learning
In view of the huge hidden dangers brought by more and more different types of malicious traffic to network securi-ty,building large-scale machine learning systems is complex and expensive,and there is less research on how to quickly build mod-els in specific scenarios in China.This paper proposes a method based on MobileNet-V2 model,which uses migration learning tech-nology to quickly build an abnormal traffic detection model.First,based on the MobileNet-V2 model,the abnormal traffic model is constructed by means of three-channel transformation and zero filling using the method of transfer learning,so as to make it conform to the application scenario of actual traffic anomaly detection and classification.Secondly,the data set adopts the USTC-TFC2016 public traffic data set,converts it into a data format similar to two-dimensional pictures through preprocessing,and inputs it into the built model for training and testing.The experimental results show that the model has good detection performance,and performs well in accuracy,precision,recall,F1 score and other main performance indicators.It can provide an efficient traffic detection scheme for other embedded devices such as firewalls.
万方数据
维普
