Research And Implementation Of Identity Authentication Based On Mobile User Behavior Analysis

With the rapid advancement and commercialization of smartphones,smartphones are becoming ubiquitous in people's lives.As people enjoy the convenience of intelligent devices and their services,they are generating a lot of mobile usage traces on a daily basis.Analyzing these usage traces,we can find out different mobile device user behaviors in order to make a distinguishing and to provide better authentication services for mobile devices and applications.In this paper,different gestures data is collected by using the sensor of the mobile device,in order to obtain the dynamic gesture behaviors.Furthermore,the authentication features of the users are extracted from the collected data to provide identity authentication based on mobile user behavior.At present,the scholars use built-in sensors of the mobile phones to carry on the research of identity authentication based on mobile user behavior,by extracting the user behavior characteristics from collected sensors data.However,the existing research proposals uses single sensors for data collection and most of them are based on acceleration sensor which is not a justified approach.The techniques based on a single sensor have a low accuracy in terms of user authentication and the attack ratio is high.At the same time,there is no reasonable update scheme for the authentication rate which is getting lower and lower due to the change of user's actions in the long-term authentication process.To solve these problem,we have proposed an efficient identity authentication method based on mobile user behavior analysis.The main research contributions and achievements of this study are as follows:(1)An identity authentication method based on acceleration sensor,gravity sensor and gyroscope sensor is proposed.Experimental results show that by using the proposed method the authentication rate is high and the attack rate is lower than that of using a single sensor.(2)A threshold method for the detection of dynamic gesture endpoint is proposed.This method is more accurate than manual control interception method,and has less computational complexity than the entropy function based interception method.At the same time,it has good interception effect for acceleration sensor,gravity sensor and gyroscope sensor.(3)For each user to build their own dynamic gesture template library,gesture authentication is based on a dynamic gesture template.The corresponding eigenvalue distance between the template library and the currently authenticated template is calculated and a new eigenvalue gesture template is constructed.Then the XGBoost algorithm is used to fit and authenticate the constructed gesture template.This scheme can make the template library update do not need to re-train the model.Finally,the obtained result is then used to judge whether the authentication is successful or not.(4)This study also provides a frequent template library update strategy to cope with the problem of the randomness in the human behavior which in turn results in the decline of the authentication rate.Based on the DBSCAN algorithm,the cluster sets with similarity can be quickly found to update template library.(5)The identity authentication system based on the mobile terminal is implemented in this study using the proposed research method.The identity authentication system can recognize the user's identity authentication after gesture.The system can also recognize the user's authentication gesture data offline by using the trained model.