The Research Of Intrusion Detection Based On Bayesian Network And Rough Set

With the development of the Internet, it has not only brought enjoy to us,and alsobrought a variety of security issues recent years. Intrusion attack is one of theserious security problems.Intrusion detection is a security measure based on the traditional security defensesystem.It provides a real-time defense for various operations.It can protect thenetwork effectively and can make it before the hazards will be able to detect theintrusion behavior. Intrusion detection is a kind of active safety protection technology.With the rapid development of Internet, the network scale is growing, the environmentof network is more complex.The intrusion detection system need to improve theprocessing speed and accuracy of detection.Among various kinds of intrusion detection algorithm, the bayesian classificationalgorithm is a kind of efficient and fast data mining classification algorithms.It is oneof the most important classification algorithm. Recent years, data mining based onbayesian network has obtained the good effect, it becomes a hot research topic, and ithas become an important research direction of intrusion detection technology.Traditional bayesian network classification algorithm based on the premise of thecomplete data, but in reality, the interception of data from the network tends to have alot of lost data for various reasons. Such traditional bayesian network learningalgorithm will not apply to the intrusion detection system with missing data. Thispaper put forward some solution to this problem. First useing of attribute reductionalgorithm based on the rough set for missing data with attribute reduction.In this paper,an algorithm of attribute reduction based on rough set is introduced. The algorithmcan effectively reduce the dimension of attributes and accurately determine thenetwork structure using DBNI with the method of distribution in Bayesian networkstructure learning from incomplete data.The attribute reduction method based on rough sets with missing data have verygood attribute reduction effect, it can effectively get rid of the redundant attributes and greatly reduce the computational complexity of classifier. At the same time, Bayesiannetwork structure learning algorithm based on the distribution distribute the lostattributes qrequency to the relevant observation frequency.It can make full use of theinormation contained in the incomplete data set. Experimental results show that thebayesian network will have a higher speed and accuracy.