Intrusion Detection Model Based On Hybrid Neural Network

With the rapid development of computer and Internet, It is gradually recognized by people that it is not enough to construct a security system only from defensive view. It is unable to meet network security needs just using firewall, digital encryption and other traditional security measures. Intrusion detection as a proactive security protection technology has become a part of the future network security study. In order to meet the needs of various network attacks, intrusion detection system develops gradually toward distributed and intelligent direction.First, the paper summarizes the current situation of network security and intrusion detection system and introduces the relevant definitions, classification and developmental trend of intrusion detection system. Secondly, this article outlines the related basics knowledge of neural network and introduces the SOM and BP network briefly. It is pointed out that the detection technology based on BP network can't adapt to the training of the large-scale samples and coding complexity. At the same time the clustering results of the network detection technology based on SOM network is not accurate enough and the new type of attacks can't be discovered. Finally, an intrusion detection system model based on hybrid neural network is presented. The BP neural network combines with the SOM neural networks to form a intrusion detection learning model with relatively complete function. The training samples decrease through data standardization and normalization, and the best matching neuron is produced using SOM neural network clustering. Then these samples with the typical characteristics are sent into the BP neural network for training. This method reduces the BP network training samples to improve the network learning speed, but also can find a good way to check the new types of attacks.The algorithm proposed in this paper realizes the corresponding module and validates the performance in the KDDCUP99 intrusion detection data set. Compared to the traditional intrusion detection algorithms, this detection algorithm can get the advantages of BP network and SOM networks and better.