Support Vector Machine Algorithm And Its Application To Intrusion Detection

Support vector machine is an important machine learning method, which originates from the statistical learning theory of structural risk minimization principle. And this method has a good performance of generalization. So, the support vector machines have attracted the attention of scholars which put forward a lot of different support vector machine algorithms. However, this method has some sensitive to noise and the training is easy to cause over-fitting problem. In order to solve these problems, by considering the different effects of each sample, researchers present the fuzzy support vector machine in order to reduce the influence of noise. In addition, in order to reduce the calculation time and further to improve the generalization performance of the algorithm, the researchers propose a nonparallel plane classifier based on neighbor support vector machine theory in order to improve the generalization ability of SVM. On the other hand, with the wide use of computer network and the rapid growth of information transmission between networks, some data for many agencies and departments has been damaged to varying degrees and the data's security is threatened. Therefore, how to prevent the intrusion of hackers has become an important security problem. At present, support vector machine is a machine learning method which is used in intrusion detection. As there exist some flaws with the traditional support vector machine, the effect using this method for intrusion detection is not obvious. In view of this situation, support vector machine algorithm needs to be further studied in order to improve the performance of intrusion detection system.Based on the idea of fuzzy support vector machine and aiming at v-SVM and twin support vector machine, fuzzy support vector machine and fuzzy twin support vector machine are studied. In addition, the presented algorithms are applied to intrusion detection. The specific contents are as follows:1. Fuzzy method is introduced into the v-SVM, fuzzy support vector machine algorithm v-FSVM is obtained. Aiming at quadratic programming's duality problem of support vector machine, iterative method for solving the problem is given using over-relaxation iteration technique.2. By introducing fuzzy method into the twin support vector machine v-TSVM and weighting each sample using fuzzy membership degree, quadratic programming problem for twin support vector machine is given. The two non-parallel classification's plane is obtained by solving the optimization problem. On the basis them, twin support vector machine algorithm v-FTSVM based on fuzzy weighting is proposed in order to reduce the influence of noise on the classification surface. At the same time, the iterative method solving v-FTSVM also is studied.3. Aiming at the presented algorithms v-FSVM and v-FTSVM, experiments are studied by choosing the standard data sets and the intrusion detection data set. Meanwhile, performance of experiment with presented algorithms is compared with the traditional support vector machine, fuzzy support vector machine and twin support vector machine for accuracy of classification, detection rate, and etc.. In addition, the iterative method is studied to further show the effectiveness of the proposed algorithms.