Study On Intrusion Detection Technique Based On Rough Sets And Neural Networks

With the abroad applications of the Internet, network intrusion and attacks increase rapidly. The issue of Computer network security becomes more and more severely. IDS (Intrusion Detection System) as a dynamic security prevention technology play an important role in protection system security. Although the existing IDS can detect the most of attacks, but with the rapid proliferating of network flux the audit data is also increase rapidly. In order to pick up representative characteristic of system mode from a great deal of audit data and construct a intelligent and efficient IDS, this paper adopt a creative IDS based on the rough sets theory and neural network. The method full use rough sets theory as a new mathematical tool to uncertain and vague data analysis has good data mining capability and advantages of neural network, such as learning from itself, association memory and identification unknown intrusion type.In this paper the rough sets theory is used to KDDCUP'99 data set then the reduce attribute set of KDDCUP'99 is used as input of neural network. The computer simulation experiments proved that the IDS based on rough sets is better than any others IDS on detection ratio and misreporting rate. When the training data up to 20% of the total data the minimum detection rate and the maximum misreporting rate attain to 98.80% and 0.77% respectively. IDS based on rough set resolved a difficult problem on how to choose characters when use neural network in intrusion detection system. The simulation experiments in this paper proved the identification system based on neural network has accuracy judgment ability when the quantity and ratio of training network's data is selected properly, In which the attack distinguish rate can up to 99.63% like DOS. So the corresponding taking-precautionary measures of network security may be adopted to aim directly at detected out intrusion action. This is of practically certain value for improving security of the computer communication network.