The Research Of Network Anomaly Detection Based On Data Mining

The21st century information revolution has enormous changed the humansociety, and changed the way of the global communication for information. With thedevelopment of computer and communication technology, network has become animportant factor in world development. Network environment has becomeincreasingly complex, so network security has become increasingly prominent.Traditional security technologies can not meet the development of computer networks.Therefore, the technology of network security has important social significance andpractical significance.Intrusion detection is popular in network security technology, which it has beenachieved certain development, and also introduced a number of commercial products.But because of high detection efficiency, poor adaptability, low extension, and othershortcomings, it is impossible to completely meet the security needs of computernetworks. It is need to continue research so that intrusion detection system is evenbetter. Based on data mining, this paper design and implement a new data miningmodel, it’s applied to network intrusion detection, improved deficiencies of existingdetection algorithms and models, improve performance of the intrusion detectionsystem. This major work done as follows:Firstly, review current development of computer network, knowledge ofintrusion detection technology and data mining, analyzes the problem of traditionalintrusion detection technology, point out that the development trend of intrusiondetection, lead to Intrusion Detection System based on data mining. Data miningtechnology improve the detection efficiency, and enhance the system’s adaptabilityand scalability.Then, depth analysis the status of intrusion detection algorithm based on datamining, proposed network anomaly detection algorithm based on anisotropiccentroidal Voronoi diagram, and construct distributed model of network anomalydetection system based on new algorithm. Theoretical analysis shows that the newalgorithm can cluster large data sets, process clustering with complex shapes,eliminate the effect of clustering brings by outlier data and "noise ", obtain theoptimal clustering.Finally, the specific implementation of new anomaly detection algorithm is described, and the system performance is detected through the data provided by KDDCup1999. The algorithm is comparable with some existing methods. Simulationresults show that the new algorithm has high detection rate and low false detectionrate.