Comparison Research Of Radial Basis Function Neural Network In Intrusion Detection

People's daily life and work increasingly depend on computer network and information technology because their quick developments, and also the network security demands reach higher level. But, even there are plenty of the current protection technologies of network security, such as firewall, access controlling and data encryption and so on, all these technologies can not entirely insure the network security and prevent the hacker's attack. So, researching and developing new intrusion detection system, which can initiatively and dynamically protect the network, unavoidly become a new primary direction of the network security.Intrusion detection technology applied many method models. A common problem within all these methods is that the the detection's rate of false positive is high and speed is is low. Furthermore, the continuously varying and changing attack modes impose more demands on flexibility and intelligency of the intrusion detection system. Neural network's outstanding abilities, parallel computing, non-linearity, self-adaptive, handling distortion data, anti-jamming, make itself a very significant tool applying in the intrusion detection field.BP model, one of the most widely used neural networks, is already applied in the intrusion detection field. In comparison with BP network model, RBF neural network is not an error back propagation method but an entire front propagation method, with shorter training time and the quality not subject to be trapped in local minimization value. At present, domestic scientists rarely study the RBF neural network and its application, especially the Probabilistic Neural Network and General Regression Neural Network. It can tell from the facts that no such papers focusing on intrusion detection fields.In this paper, three radial basis function neural network models combining with adaptive resonance theory are presented for intrusion detection classification and prediction. By rule extracting arithmetic, they increase detection rate and reduce false positive rate of an intrusion detection system, which can be used to bothmisuse detection and anomaly detection.