With the development of the Internet, network security incidents occur frequently, which have a serious impact on the country and society. In order to analyze the distribution characteristics of the extreme value of website visits, this paper takes the traffic data of UESTC logistics website as the research object, and constructs a generalized extreme value model. First of all, we make a preliminary statistics on the website visit data, which reveals that the extreme visit is not common. Secondly, the parameters of the generalized extreme value model are estimated by the maximum likelihood estimation, and the model diagnosis shows that the generalized extreme value model has a good fitting effect. Then, the return level of 1 month, 2 month, 3 month return period is analyzed, and it is proved that the generalized extreme value model has a certain extrapolation ability, and can calculate the return level of a long return period. Finally, the fitting effect of the Generalized Extreme Value model is analyzed when the earlier historical data is discarded, and it is proved that the Generalized Extreme Value model is not sensitive to the longer historical data, and the model has good stability.