Research On SDN Anomaly Detection Based On Deep Learning

With the rapid development of the information age,more and more network users,the traditional network architecture can not meet the growing network traffic,seriously affecting the user's online experience.Therefore,there is an urgent need for a network architecture that can monitor network traffic in real time.The proposal of Software Defined Networking(SDN)provides an effective solution for the problems encountered in the current network,and is generally regarded by the industry as the development direction of the next generation Internet.The core of the SDN network architecture is the OpenFlow protocol mechanism,which is an open protocol standard.However,this mechanism is prone to network attacks.How to detect network attacks is the key to SDN network security.In this context,this thesis proposes a GRU-CNN SDN anomaly detection network architecture to enhance the reliability of SDN and improve the service quality of the network.Firstly,the thesis expounds the research significance and current situation of SDN anomaly detection,focuses on the advantages of SDN network architecture,designs an SDN anomaly detection architecture,and analyzes and studies the operation mechanism of SDN anomaly detection architecture in detail.The SDN anomaly detection architecture mainly includes three modules: traffic collection module,anomaly detection module and data transmission module,and fully researches and explores the functions and functions of each module and the interconnection between modules.Abnormal traffic detection is essentially a process of classification,and deep learning has a natural advantage in classification.Gated Recurrent Unit(GRU)has achieved the best classification in timing characteristics,but there is still a lack of spatial domain;and Convolutional Neural Network(CNN)can be obtained in the spatial domain.effective.Therefore,this paper proposes a model for applying GRUCNN to SDN for anomaly detection.In order to accelerate the convergence of the model and prevent the model from over-fitting,the model optimization based on the Adam optimizer and the regularization method is proposed to further accelerate the convergence of the model and prevent the model from over-fitting.Finally,the NSLKDD data set uses the GRU-CNN model for anomaly detection,and its accuracy rate reaches 97.74%,which provides a guarantee for the secure transmission of SDN network information.