A Secure Authentication Method Of Intelligent Terminals Based On JS-Divergence

With the development of intelligent terminal,people have been used to dealing with a variety of transactions on it which stored more and more user privacy information.Therefore,security authentication of intelligent terminal is particularly important,since people faced with its increasingly serious security issues when enjoying the convenience brought by intelligent terminals.The traditional explicit authentication mechanism,such as PIN code,password,gesture patterns,is easy to use,but there are still some flaws.First of all,it is vulnerable to shoulder surfing,which makes the leakage of authentication information and leads to lower security.Secondly,users need to enter the authentication information frequently owing to the high frequency and short-term of usage characteristics of terminal,which causes poor usability.To end these,the implicit authentication mechanism that typically uses various types of behavioral characteristics to achieve the authentication of the user identity came into being,which does not need users' explicit action and potentially enhances the experience of users and further ensures security.However,the implicit authentication is in the development stage,there are still some shortcomings,such as computational instability,difficult to achieve and so on.In order to solve the above problems of intelligent terminal security authentication,the ultimate goal of this paper is to propose an intelligent terminal security authentication method with stable computation,good user experience and high security.Combining with the user behavior characteristics,this paper firstly proposes a characteristics similarity calculation method based on JS divergence which has computational stability and can be used to calculate the similarity between probability distributions of random variables.And its feasibility is proved by organizing the students in a class and collecting their behavior characteristics.Then,an intelligent terminal security authentication method based on JS divergence is proposed,which uses the proposed similarity calculation method and the fact that JS divergence is rarely used in the field of intelligent terminals authentication.This method is divided into the user training phase and the user authentication phase according to the FIDO authentication framework,which achieves user authentication through preprocessing and reconstructing the behavior characteristics sequence collected when the user enter the password for authentication by touch screen.The proposed authentication method does not require the user input behavior characteristics deliberately,so as to have a good user experience.And it re-trains the user behavior characteristics model by updating the training set using the authenticated behavior characteristics to further improve the security.Considering the limitation of single factor authentication,the proposed authentication method adopts two-factor authentication including the authentication password and the user behavior characteristic as a secondly authentication factor.Particularly,the behavior characteristics of the proposed method can be a variety of numerical characteristics and any combination of them.A series of experiments were made on two datasets to evaluate the performance of the proposed authentication method.In the experiments,we compare and analyze the impacts on the authentication results from four aspects: the size of the training sample set,the number of interval division,the re-training and the different distance measurement methods,respectively.The results are presented using FRR,FAR and ERR,which tell us that the proposed authentication method based on JS divergence is better than that based on other distance measurement methods by choosing the appropriate training sample set and the number of interval division.