Research And Implementation Of Passwordless Authentication Technology

At present,the Internet is closely related to people's lives,and network security incidents are also emerging.People's privacy data on the Internet has become the target of attackers.Identity authentication is the first link in the login system and the first line of defense to protect network security.Traditional static password-based authentication methods require human memory and weak passwords.They are also vulnerable to attacks such as brute force attacks and database collisions.Therefore,passwordless authentication technology has become a research hotspot at home and abroad.The existing passwordless authentication schemes mainly use hardware devices and biometrics for identity authentication.However,intelligent hardware-based authentication methods face hardware damage,loss,or theft Risk,and the authentication method based on the user's physiological behavior requires additional reliance on professional information collection equipment and can only provide one-time authentication services,and cannot perform continuous behavior detection on the operation after the user logs in.In view of the above problems,this topic studies the technology of passwordless identity authentication.First,the software operation behavior of the user terminal system and the browsing behavior of the WEB terminal are used to separately train a user behavior classification model based on deep learning,and the trained deep learning model is used as a user behavior feature matching template.At the same time,in order to avoid the problem of easy failure of a single authentication mechanism,a multi-model data fusion method is adopted to fuse the two behavior classification models to establish a user dual-view behavior authentication model.Secondly,in multi-model data fusion,a weighted adaptive weighting and fusion algorithm is proposed to tune the fusion weight of each part of the model to obtain the optimal fusion result,and to avoid the effect of artificially setting the weight on the fusion result.Further improve the recognition accuracy and robustness of the model.In addition,the algorithm can also dynamically adjust the fusion weight of each sub-model to adapt to the authentication model when a certain behavior habit of the user changes,and assign higher fusion weight to the high recognition accuracy model,thereby improving the authentication accuracy rate.Finally,this paper uses a trust management model and combines the two innovative and improved algorithms to design and implement a passwordless authentication algorithm library for third-party calls.This algorithm library can effectively provide users with PC-based operating behaviors and web browsing behaviors.Identityauthentication services,as well as services that continue to dynamically detect the authenticity of users' identity after login,and prove the effectiveness of the algorithm through experimental tests.