Na (?) Ve Bayesian And Svm In The Detection Of Ddos Attacks

With the wide application of Internet and the development of network techniques, people contact with network closely. And network has been an important means of getting and transmitting information for people. More and more Network security happen frequently, because the number of Internet users increase quickly and the network is widely used in every fields. These network security events make the social production and people's life insecurity and negative effects. Therefore, we should be attention to network security,whether the national view or a personal view.Among all the network security problems, Distributed Denial of Service(DDoS) attacks have been harmful and difficult to resolve. DDoS attack is an attack that the attackers use various of tools and techniques to make the Internet services or computer systems can not work and provide services for users.Firstly, this paper analyzes the background of DDoS attacks, the principle of attacks,means of attacks and defense testing techniques. And through the study of machine learning methods which is a means of detecting DDoS attack, this paper use the Naive Bayes theory and support vector machine theory in the DDoS attack detection. The main work includes:(1) Naive Bayes theory is used in the DDoS attack detection. This paper design a model Naive Bayes DDoS Attack Detection Model (NBAD), which detect attack quickly. And the model is a self-learning model.NBAD analysis method to solve the model and architecture model algorithm is given NBAD pseudo-code description.(2) Support vector machine theory is used in the DDoS attack detection. This paper design a model:SVM based DDoS Attack Detection Model (SVMAD), and analyze the basic principles of SVM and the key issue in DDoS attack detection(3) The experimental platform is the Weka,which is a very famous data mining tool. The data set is KDD CUP 1999 for the experiment, using the 10-folder cross-validation to construct an environment.(4) According to the classified predict experiment of NBAD model and SVMAD model, it proves the validity of the models. A classic machine learning decision tree algorithm J48 were compared with NBAD model and SVMAD model, which are analyzed detection performance.