Research Of Intrusion Detection Model Based On Rough Set And Neural Network

With the development of technology and network scope, internet is facing more and more intrusion and danger. The issue of internet security became more serious. As an important chain of internet security resolvent, intrusion detection technology has got full attention.Facts prove that traditional computer security theory already couldn't fit the development and movement of internet environment. Traditional intrusion detection systems have many shortcomings about validity, adjustability and expansibility. So neural network, BP algorithm and rough set theory are introduced into the field of intrusion detection for improving the performance of intrusion detection.The paper mainly discusses some effective means that can improve the ability of intrusion detection. The thesis begins its discussion by introducing the concept of IDS, the classification of the IDS, rough set theory and neural network. Then we analyze the characteristic and defection of intrusion detection technology, combine the respective superiority of rough set and neural network, and propose an Rough Set and Neural Network Intrusion Detection Model (RS-NNIDM). In this model, we make the rough set as a prepositive disposing system to reduce the attributes. Through the optimal attribute reducing algorithm based on cognizable matrix, the least attributes group is obtained which can distinguish intrusion from normal behavior. And the intrusion characteristic database of expert system is set up to detect known intrusion. Then, the structure of LVQ neural network is also set up according to the amount of least attributes group, the learning swatch set is simplified by rough set and is used to train the neural network. For some unsure class or unknown intrusion of expert system, we can use neural network to classify them and update the characteristic database at the same time.For the function of every module of our model, we make detailed analysis, design and way of implementation. The expert can detect not only known intrusion behaviors but also unknown intrusion by retraining the neural network.We carried on the relevant experiment to proof-test the performance of RS-NNIDM. The experiment result shows that this method has overcome the limitation of BP which need more time to train and has a lower constringency rate. So it's prior of other method. This article is an attempt which we applies rough set and neural network to ID. The model plots out modules according CIDF and has great effect on both theory and practice.