With the development of information technology,Website is indispensable in internet applications.Web development and network security are key issue for modern management and international market competition,so Web development and network security are equally importment for site structure.Intrusion detection is an important area in network security.At present,a lot of technologies are applied to this field.such as Intrusion detection based expert systeim,Intrusion detection based neural network,Intrusion detection based reasoning and so on.Compared with these technologies,Anomy intrusion detection technology based on outlier mining has some advantages:it has no training process,so overcomes high false alarm rate which due to incompleteness of finite training sample;compared with the whole network, Intrusion detection is viewed as isolated point of data sets or a small number of anomaly detection data.Our research is based on Northeast large-scale scientific instrument information sharing network.B/S three architecture is used.Taking into account system performance and data security,C/S architecture is used in system because of maintenance and patch date implementation.Component Interface Pattern is.NET Framework.The process of Anomy intrusion detection for outlier mining based on similar coefficient sum are as follow:Input data sets,Data preprocessing,Calculating similarity,Calculating the sum of similar coefficient and output outlier mining sets.Our forty experimental datas come from Northeast large-scale scientific instrument information sharing network,C++is used,The algorithm we proposed divide connection records into three patterns,The experimental results show that Anomy intrusion detection for outlier mining based on similar coefficient sum not only improves detection rate of invasion attack,but also more effective than K-means. |