Research On Mobile Phone Continuous Identity Authentication Based On Behavior Characteristics

Mobile phone has become an important equipment for communication,entertainment and payment,especially as a means of payment instead of cash,the importance of maintaining its security is self-evident.The existing main methods for mobile phone security include lock screen pattern and lock screen password,and payment software has payment password.However,once the mobile phone is lost or the password is leaked(for example,some mobile phone users like to write down the password on notebooks and small notes;When the user is using the mobile phone,the non authenticated user next to the user can directly see the sliding gesture pattern or the password that the user uses when unlocking the mobile phone).The non authenticated user who obtains the mobile phone or the password will pose a great threat to the security of the mobile phone.The mobile phone continuous identity authentication can continue to protect the security of the mobile phone when the password of the non authenticated user is known.Continuous identity authentication is an implicit process that uses the resources of mobile devices and built-in sensors to continuously or regularly capture the behavior attributes of current mobile phone users,and verify whether they are authenticated users by analyzing these behavior attributes.In recent years,more and more researchers are interested in the field of continuous identity authentication.Especially with the expansion of storage and computing resources and the availability of sensors,continuous identity authentication becomes more and more accurate and effective.In this paper,a continuous authentication method based on user touch feature map is proposed.By collecting the touch position,pressure,area and sensor data generated by the user when touching the screen,these data are processed and written into pictures: a single touch feature map reflecting the user's one touch feature,the interval feature map reflecting the user's continuous input features and the sliding gesture feature map reflecting the sliding unlocking scene and the long touch scene features.In this way,the recognition of users turns to the recognition of image field.In this paper,convolution neural network and several common image binary classification algorithms are used to test the above three kinds of feature images,and convolution neural network gets the best classification effect.This paper discusses the classification effect of recognition strategy using different number of single touch feature maps and combined with interval feature maps.In the case of using six single touch feature maps and their corresponding interval feature maps,the recognition effect of error rejection rate(FRR)of 0.34% and error acceptance rate(far)of 0.25%can be obtained.The influence of mobile state,one handed operation state and a small amount of training data of authenticated users on the recognition effect is extended.This paper discusses the impact of using different number of sensors on the classification effect,and builds a continuous identity authentication model,which can identify the user's identity with a response time of 1.9 seconds after the user touches the screen.