Research And Implementation Of Identity Authentication Mechanism Based On Smart Phone

With the rapid development of the information age,smartphones are no longer simply used as a communication tool,but also as a social media store of a large amount of personal information and privacy.Security is particularly important.Traditional mobile phone authentication methods such as Pin codes,alphanumeric passwords,hand-painted grids,and password information may be leaked;biophysical characteristics such as face recognition and fingerprint unlocking also risk being imitated or forged by intruders.There are a large number of built-in sensors in smartphone,such as accelerometer,gyroscope,etc.The data they collect contains gestures,behavioral time-frequency characteristics,and specific behavioral habits of different users.Using these features can effectively authenticate the user of the mobile phone,and it is not easy to be imitated.In this thesis,specific gesture and behavior information collected by the built-in accelerometer and gyroscope sensors of the smart phone are used to realize the identity authentication based on gesture recognition and continuous identity authentication based on behavior recognition.In the gesture authentication-based identity authentication,the collected gyroscope data that writes 3D gestures in the air with a handshake is preprocessed and its effective time-domain features are extracted,and the isolated forest algorithm is proposed to verify the validity of specific gesture data.On this basis,the bi-directional long short-term memory recurrent neural network(Bi-directional Long Short-Term Memory,BiLSTM)is used to effectively recognize gestures and achieve secondary authentication.In identity authentication based on behavior recognition,the accelerometer and gyroscope data collected for six types of behaviors of normal walking,fast walking,slow walking,running,going up and down stairs are pre-processed such as gait cycle detection and division.Using BiLSTM to effectively identify behaviors.This paper proposes a network model of Convolutional Neural Networks(CNN)combined with bidirectional long-term and short-term memory recurrent neural network.It uses convolutional neural network for feature extraction and two-way recurrent neural network for binary classification to achieve identity in a specific behavior mode.Experimental results show that in gesture-based authentication,the averageaccuracy of gesture recognition is 94.3%,which supports custom gesture recognition.The average accuracy of gesture-based authentication is 88.3%,while the average accuracy of gesture integrated authentication is 83%.In the identity authentication based on behavior recognition,the average accuracy of behavior recognition is 92.5%,which can recognize similar behaviors.The average accuracy of behavior-based authentication is 90.2%,and the average accuracy of behavior comprehensive authentication is 84.4%.The experiments also compared the effects of different numbers of legal users,different gesture writing methods,and different walking methods on the accuracy of identity authentication.Experiments show that the method proposed in this thesis is feasible in solving smart phone authentication.Gesture recognition,behavior recognition,and gesture or behavior-based authentication can achieve high accuracy.