Research On Continuous Authentication Of Gait Behavior For Smartphones

With the development of the mobile Internet and the popularization of smart phones,the user privacy data contained in mobile phones has increased,and how to ensure user privacy and security has become a research hotspot in academia and business circles.Traditional authentication methods,such as passwords,facial recognition,and pattern unlocking,only perform one-time authentication when the mobile phone is initially logged in,and cannot be continuously authenticated during user access,giving illegal users an opportunity.Continuous authentication based on user behavior provides a solution,which can collect behavior data in a way that users do not perceive,and continuously determine the legitimacy of user identity,thereby ensuring the privacy and security of smartphones during access.In recent years,gait as a basic behavior of the human body has been widely used in continuous authentication,and gait authentication based on built-in sensors in smartphones has received extensive attention.However,most of the existing studies have the following defects: first,they fail to meet the needs of high-accuracy authentication in multi-gait scenarios.Here,multi-scenarios specifically refer to different gait activities and different mobile phone placement locations;second,Gait model training and user authentication are difficult to take into account high efficiency and high accuracy;third,the spatiotemporal trajectory information of gait behavior is not considered,and the stability and security of the authentication system need to be improved.Therefore,this paper designs a GRU-based gait authentication scheme for defects one and two,and designs a gait authentication scheme integrating spatiotemporal trajectories for defect three.The specific research contents of this paper are as follows:(1)In order to achieve high-accuracy authentication in multi-gait scenarios and efficient training of user models,a GRU-based gait authentication scheme is proposed.First,noise reduction and gait cycle segmentation are performed on the gait signal obtained by the mobile phone motion sensor.Secondly,in order to meet the authentication requirements of multi-gait scenarios,the coordinate system transformation method is used to perform direction-independent processing of gait signals.Third,in order to achieve high-accuracy authentication and high-efficiency training of the authentication model,this paper uses gated recurrent units(GRUs)with different architectures and various optimization methods to train the user gait model.In the authentication stage,a reward and punishment authentication mechanism based on trust value is used to judge the legitimacy of the user’s identity.Finally,experiments are designed to evaluate the performance of the scheme.The experimental results show that compared with the listed schemes,this paper improves the authentication accuracy in different temporal scenarios,and the Equal Error Rate(EER)of authentication on two public datasets is 3.94% and 4.51%,respectively.Compared with the gait authentication model based on Long Short-Term Memory(LSTM),the training efficiency of this working model is improved by about 20%,and the memory usage is smaller.(2)To cope with the limitation of single-behavior modal authentication of motion sensors,a gait authentication scheme based on fusion of spatiotemporal trajectories is proposed.First,in order to realize the stereo positioning of the user plane and vertical direction,this work analyzes and processes the spatiotemporal information of the user’s gait behavior obtained from GPS and Wi Fi data,divides it into multiple spatiotemporal regions,and extracts features for training spatiotemporal trajectories Authentication model.Secondly,in order to flexibly integrate different modalities,this work chooses to perform fractional or decisionlevel fusion of gait authentication and spatiotemporal trajectory authentication results to achieve fusion authentication.Finally,experiments are designed to evaluate the spatiotemporal trajectory authentication scheme and the fusion authentication scheme respectively.The experimental results show that the gait authentication based on the fusion of spatio-temporal trajectories improves the performance of the authentication system,specifically,its authentication accuracy is higher,its ability to resist adversary attacks is stronger,and its system stability is stronger.