Application Of CAPTCHA With Behavioral Valification Based On GBDT

Currently, in our society the information developing rapidly, the Internet has allready penetrated into our lives everywhere, become inseparable part of our lives, which not only bring us so much convenience, but also many potential problems of network security issues. As the capillaries of network, identifying code naturally been given to maintain the network security.Firstly, for identifying code widespearding in our dailylife, this paper based, introduces the various types of identifying code and their characteristics, the important roles of identifying code in the online world, and the most popular style of behavior codes currently, through development of the identifying code technology. This identifying code has high demands to the CAPTCHA model.Secondly, briefly introduces the basic methods and the advantages and disadvantages naive Bayes, logistic regression and decision tree classification model, focusing on the most leading-edge machine learning thinking relation to the ensemble learning model. Gradient Boosting Decision Tree Model, an achievement of ensemble learing, which using the decision tree as the bottom learning model.Finally, extracting data from authentication sliding track data of the behavior-based verification code, using the boosting gradient classification tree algorithm to classify the behavior of human and machine,to distinguish humans and machines. Finally, respectively, training the naive Bayes, logistic regression, support vector machine classification algorithm comparing their results, we find that the gradient boosting decision tree model based on gradient Boosting method has the best performance.