Research On Continuous Identity Authentication Based On Non-reproducible Multi-factor

User authentication of computer systems can be seen everywhere in daily life.Most systems use traditional one-time authentication methods such as passwords,fingerprints and facial recognition based on biometrics.But passwords are at risk of being cracked or forgotten,and both fingerprint and face recognition require specific hardware devices.For the one-time authentication method,once the intruder passes the authentication,The operation of the intruder after entering the system is no longer restricted.Operational activities in computer system can be regarded as a biological behavior characteristic,which cannot be stolen or copied,and can be used for identity authentication.If the user's operation behavior(keystroke,mouse click and slide,etc.)is abstracted into a behavior model,the application model continuously judges the user's identity to achieve continuous authentication.Continuous authentication overcomes the disadvantages of one-time authentication.The user will generate a large amount of static environment information when operating the system.The purpose of uniquely identifying the host of the system can be achieved by integrating useful static information.This paper studies the continuous authentication of the multi-factor behavior model from three aspects:user keystroke behavior,mouse click and slide behavior,and host static information.The main research work and results are as follows:1.Free-text keystroke identity continuous authentication based on deep learning model:Construct a deep learning model of convolutional neural networks(CNN)plus recurrent neural networks(RNN).The application model is used to train user keystroke data to obtain a user keystroke behavior model.Using the trained model to test the optimal equal error rate(EER)of 3.04%on the open data set.Combining the model with a sliding window,the decision algorithm implements a prototype of a continuous identity authentication system based on keystroke behavior.2.Human-computer recognition model based on mouse operation behavior:In-depth analysis of advanced features of mouse click and slide,and using machine learning algorithms to train mouse operation behavior model to identify machine behavior.Auxiliary model in continuous identity authentication.3.Host discriminant model based on host static information model:Research on multi-dimensional,multi-level static information that can identify user host.Integrate the available statics,and use the comparison algorithm to build a host fingerprint model to identify users' commonly used hosts.