Bp Neural Network-based Intelligent Intrusion Detection System Design And Realization

Internet has been an important part of people's lives, and it will greatly change the information and other important industries. However, because of the opening and frangibility of network, the bug of OS, software and hardware, there are a lot of challenges and dangers in network security now. Recently, more and more network attacks have brought enormous loss. Network attacks have become more complex, flexible and extensive than ever before. A lot of secure policies, strategies and tools are analyzed and applied to solve these problems, including IDS (Intrusion Dection System). IDS can provide secure strategy in time based on an analysis of attacking actions. But common IDS can't identify unknown attacking actions; it needs people to analyze a lot of network data. There isn't a mature intelligent IDS product yet, which can identify unknown attacking actions automatically. To solve the above problems, in this thesis, the author researched the techniques for IDS and neural networks, designed a kind of intelligent IDS based on BP neural networks, proposed a creative system structure about five iterative BP neural networks, and gave the methods of transforming network data into mathematic data of each BP neural network.The author completed detailed design and finished most key codes. Five BP neural networks were trained and tested by a lot of network data including multifarious attacking actions. The test result proves that this intelligent IDS can achieve better performance than corresponding traditional ones and can detect some unknown attacking actions. An analysis of the insufficiency of this system was given in the end of this thesis. The author brought forward the method to improve on this system and pointed out the next task of this system.