Study On The Application Of The Intrusion Detection System Based On The Artificial Neural Network

We have proposed an intrusion detection system prototype based on the framework of the Common Intrusion Detection System and the software packages; we have described the functions of the modules of the prototype,discussed the relations, the methods of the realization and the mechanism of the modules. We have proposed an easy way to transfer the isomerous data to isomorphic data, to unitary eigenvector to adapt to the requirement of the artificial neural networks.The objectives of the paper are: how the artificial neural network plays in the application of the intrusion detection? Whether the artificial neural network can identify the unknown attacks?Because the BP neural network plays very well in many applications, having the good adaptation and generalization ability, we adopted the BP network to carry on our research works.We have validated that the artificial neural network can identify the unknown attacks and plays well in the application of the intrusion detection except the paroxysmal attacks which mode the training data set has no mode to resemble and which numbers are very little.We found that the BP neural network may train very slowly when the training data set is very big, which might seriously affect the performance of the BP network, so in later researches we should adopt the other techniques of the intrusion detection such as data mining, the rule and other artificial neural network algorithms.Although the mechanism of the artificial neural network is parallel, it still runs in serial mode on the Von Neumman machine, which seriously affects the applying result of the artificial neural network. If the performance of the computer is high enough, the more the training samples and modes are, the higher the accuracy of the artificial neural network forecasting is.