Analysis On Network Security Technology

By the development of internet,it entering every corner of the world gradually, people have deepened understanding of the Internet,that the Internet behavior increaseing by a sharp surge. Especially when we get into the 21st century, the chain of internet industry is growing by the unprecedented and rapid momentum, both live or work, it has been unable to survey without Internet. With the continual emergence of new businesses, the requirements of better internet service quality are also increasing, quality of internet service is facing a huge challenge. More and more researchers which base on existing infrastructure have devoted to the optimization of Internet management, controling of the attributes and characteristics of network flow, so that realize the optimal of network performance.Network flow is random, sudden, dynamic and real-time, meanwhile network traffic has self-similarity, not only in a short period of time show correlation with the same statistical properties,it shows long correlation of its high variability in the long-term data flow information, It's effective to evaluation the performance of network is well or not by modeling and analysis the network traffic .and it will give a good guide to network design, control and adjust with accurate predictions.The research in this paper aim to establish a new network model that could describe the characteristics of network traffic better and reach to a better predication result which more approach to the value of the traffic in the future. This paper analyzes the present status of network traffic analysis and forecasting,for prediction model of BP neural network can easily fall into local minimum,It's necessary to optimization the algorithm and establishment of a new forecasting model. This article introduces the application of wavelet packet neural network in the traffic predication,datas pretreated by wavelet packet analysis discard the influence of the noise,that made training output value more closer to the actual flow value. Simulated Annealing algorithm has a strong ability of local search,which can jump out of minimum point of the training network, reach to a global optimal result. Kalman filter algorithm could eliminate random observation noise and interference signal and extract useful informations,In the training process, the SA algorithm not only update network weights in an orderly way, but also update approximation error variance,which carry second derivative information.Combine these two algorithms, this article propose an improved wavelet packet neural network model. Through the simulation, it show that the new model is effective.The work in this paper indicate that the with improved model training has a better performance.It could fitting the traffic data well,and the accuracy of prediction is significantly better than the wavelet packet neural network model.