The Research On Sliding Verification Code Human-Machine Recognition Based On Support Vector Machine

The constant innovation and progress of science and technology has made people have to face the problem of network security while enjoying the benefits of technology.The emergence of the verification code provided the important guarantee for road network security,but with all kinds of authentication code decoded by a machine automation program continuously,not only affects the normal operation of the network system,at the same time to bring economic loss to normal users,such as information leakage hidden trouble.Simple verification codes,such as digital characters,are easy to use but easy to be deciphered by machines.Although the q&a and picture verification codes have high security performance,but the user experience is poor,so it is a good choice to use the convenient and certain security sliding verification codes.Verification code based on the sliding type,use of humans and machines to complete validation left by the sliding track to extract the feature data,combined with modern statistical method to establish model,try to solve correctly identify the man-machine operation this vital network security problem.First,we marked the machine behavior sample as the positive class,which is the class of concern,and the human behavior sample was marked as negative.Secondly,the correlation analysis of all sample data shows that the correlation between sample indicators is small and the interaction is weak.Each sample can play a role in establishing the classification model.Then,based on the standardized data,the support vector machine(SVM)algorithm in machine learning is used to establish the support vector machine classification model to predict and verify the categories of operators.At the same time,R statistical software was used to conduct empirical research and analysis,and support vector machines were compared with Fisher discrimination,decision tree and BP neural network classification algorithms.The results show that the sensitivity of all model recognition machines is higher than that of the real person,which is in line with the research purpose of this paper.The classification model of SVM is superior to other classifiers,and the classification accuracy and AUC value are higher than other classification models.The accuracy and AUC are 94.82%and 0.943 respectively.Therefore,the support vector machine classification model can meet the need of sliding verification code and distinguish the verification operation behavior between machine and human.This means that the machine is identified and rejected when it completes verification.Finally,this paper summarizes the main research contents,and puts forward the prospect for some deficiencies.