Smartphone Authentication Method Based On Keystroke Characteristics

With the continuous development of password-based brute force attacks,smudge attacks and frequent occurrence of user information leakage,password-based smartphone authentication method is no longer safe and reliable.The emergence and development of biometrics brings hope to solve this security issue.Biometrics uses user's own behavioral or physiological characteristics to perform identity authentication and identification.At present,the research of biometrics mainly focuses on physiological features such as face,voice,fingerprint and iris,but these biometrics require the support of additional hardware devices,which is not conducive to the reduction of production costs and the integrated integration of devices.The emergence of authentication technology based on keystroke dynamics makes up for this deficiencies.Therefore,this paper studies the keystroke dynamics of smartphone devices in details,proposes and implements a smartphone authentication method based on keystroke dynamics.The main work of this paper is as follows.Firstly,this paper analyze and summarize the keystroke dynamics related research.And then make a deep analyze and reasearch on keystroke dynamics and classification algorithms.Finally found that the existing reasearch has following problems.On the one hand,the first-order keystroke characteristics are associated with the user password length,and the number is relatively large,resulting in slower training and recognition of the model.The keystroke-based smartphone authentication method faces the problem of positive and negative sample data set imbalance in practical application scenarios.The binary classification algorithm has a high user identification rate but can not solve the data set imbalance problem.The one-class classification algorithm can solve the data set unbalanced problem,but the user recognition rate is low.Secondly,in order to solve above problems,this paper proposes a smartphone authentication method based on keystroke dynamics.On the one hand,this method proposes a specific method of extracting the second-order keystroke features based on the first-order keystroke features.The second-order keystroke features is more in line with the user's keystroke behavior patterns.On the other hand,this method studies the existing problems in dealing with the imbalanced data set algorithm.Based on the Tomek Liniks undersampling method and the improved SMOTE algorithm,a two-class classification algorithm based on TL-SMOTE is proposed.The experimental results show that the proposed method can not only solve the problem of unbalanced data set,but also has a high user recognition rate.Thirdly,make a system design and then implement the smartphone authentication method based on keystroke dynamics which is proposes above.The system seamlessly combines identity verification modules based on keystroke dynamic without affecting the normal operation of the old system based on username/password.The addition module is transparent and invisible to users,so users can gain enhanced protection based on keystrokes-based authentication without any additional learning costs.