A Study Of Intrusion Detection Based On The Weighted Fuzzy Support Vector Machine

The rapid development of network technology, information sharing today although increased exchanges between people, increase productivity, but also make individuals, companies and organizations, the private information even countries confidential faces serious invasion of security threats, various security problems occurred frequently, therefore the computer information security more and more receives people's extensive concern. Intrusion detection system as a kind of positive safe protection tools, not only provides for internal attack, external attacks and wrong operation in real-time protection, and computer network and system harm for alarm, intercept and before responding. Network security intrusion detection system as a firewall important supply has many shortcomings, such as high misstatement rate for new attacks cannot detect effectively, therefore will artificial intelligence learning algorithm is introduced to intrusion detection to solve these problems is the current research hotspot. At present there are many researchers will neural network algorithm is introduced to intrusion detection and achieved good effect, but the neural network has many shortcomings, such as: local minima problem, learning problems, etc in intrusion detection limit their the further development of application.This paper firstly expounds the background and meaning, the paper introduced the intrusion detection model, classification, compared the application in intrusion detection, different methods of advantages and disadvantages. And then introduced the statistical learning theory, Support Vector Machine (SVM) Support Vector Machine, with Fuzzy Support Vector machines (Fuzzy Support Vector Machine, FSVM) related theory, is proposed based on the average density of weighted Fuzzy membership tectonic methods are used to reduce outlier and noise points separating hyperplane of constructing the role of network intrusion detection, according to the characteristics of the chosen the suitable RBF kernel functions and punish value, construct network intrusion detection fit within the Fuzzy Support Vector Machine classifier and will this classifier applied to network intrusion detection, through the data of DARPA experiment result shows that the classifier used in network intrusion detection will be feasible.