Research On DDoS Attack Detection Based On Optimized RBFNN

With the continuous development of network technology and the diversification of attack means,the traditional distributed denial of service(DDoS)attack detection technology gradually shows the problems of poor detection performance and low adaptability.The new technologies and methods of DDoS attack detection have become a research hotspot in related fields.Software Definition Network(SDN)has the characteristics of centralized control.Collecting traffic for attack detection under SDN can improve the performance of DDoS attack detection.Based on this,this paper constructs an DDoS attack detection model under SDN structure by optimizing Radial Basis Function Neural Network(RBFNN).In this paper,the gradient descent algorithm suitable for DDoS attack research is selected as the training method of the parameters of RBFNN by analyzing the training methods of the parameters of RBFNN.In order to improve the optimization speed of the gradient descent algorithm,a dynamic step-size gradient descent algorithm is proposed to calculate the step-size dynamically according to the change of loss function.The experimental results show that the dynamic step-size gradient descent algorithm has fast convergence speed and strong optimization ability.The RBFNN algorithm is optimized based on the dynamic step gradient descent algorithm to optimize the parameters of the RBFNN algorithm to improve its learning ability.And And the verification experimental results indicate that the algorithm has high learing ability.Based on optimized RBFNN algorithm,aiming at the problems of poor DDoS attack detection performance and low adaptability,a DDoS attack detection model based on optimized RBFNN is proposed under SDN.The model needs to input the DDoS attack feature,The IP address entropy ratio is used as one of the DDoS attack feature values so as to distinguish the difference between the network traffic burst and the DDoS attack.Finally,the RBFNN attack detection model is tested by simulating DDoS attacks under SDN,and the model is proved to have good performance in attack Accuracy Rate,False Positive Rate and False Negative Rate.The experimental results show that the proposed RBF neural network attack detection model based on dynamic gradient descent algorithm optimization performs well in the Accuracy Rate,False Positive Rate and False Negative Rate,and the performance of eachindex is stable under various attacks.