Analysis And Application Of User Identity Authentication Technology Based On Keystroke Behavior

With the continuous advancement of industrial informatization,the importance of information security is increasing day by day.Identity authentication is an important research topic in the field of information security.In recent years,biometric authentication technology using human physiological and behavioral characteristics has been increasingly developed and applied.Compared with human physiological characteristics such as faces,fingerprints,and iris,and human behavior characteristics such as sound,stylistics,and gait,the advantages of keystroke behaviors are that the collection of keystroke data does not require high-cost hardware equipment;There is less user intervention in the authentication application process.However,there is still room for improvement in the accuracy of keystroke-based authentication.Secondly,the existing authentication technologies are mostly one-time authentication,which cannot achieve the real-time identity monitoring.In addition,the practical application of keystroke features requires many considerations such as cold start problems,registration sample features missing.This paper focuses on the issues related to identity verification and identity monitoring in the field of information security,and proposes a scheme for user identity authentication using user keystrokes.By collecting keystroke data generated by users when they use the keyboard,the analysis and modeling of user's keystroke behavior are performed.We also designs authentication strengthening methods for problems that may occur in practical applications to improve the authentication accuracy and to some extent solve the problems of cold start and missing features.Finally,static authentication based on fixed text and continuous authentication based on free text are implemented.The main research work of this paper includes:(1)Framework design of identity authentication method based on keystroke behaviorThis paper presents a framework of authentication methods based on keystroke behavior.The framework is mainly divided into the keystroke information input module,the model storage module,the static authentication module,the continuous authentication module,and the authentication enhancement module.It implements three keystroke application scenarios:login keystroke authentication,important node authentication,and continuous input behavior authentication.(2)Design of static authentication method based on Multivariate Gaussian modelThis paper designs a static authentication method based on a multivariate Gaussian model,including feature extraction and extreme value filtering of keystroke features,keystroke model construction and incremental optimization methods,keystroke sample feature decentralization fusion method,and the calculation method of threshold based on LOOM rule.Finally static authentication of user identity based on fixed text is realized,and the authentication accuracy is improved.(3)Design of continuous authentication method based on RNN prediction model and dynamic trust mechanismThis paper designs a continuous authentication method based on RNN prediction model and dynamic trust mechanism.Firstly,the keystroke data stream of free text input is split and parsed,and the user's keystroke behavior characteristics are learned by using the prediction model based on RNN.Then,based on the dynamic trust mechanism,each keystroke in the authentication process is predicted by the model and the trust value is evaluated by the model.The continuous authentication based on a single keystroke is finally realized,and the user identity can be monitored in real time in the typing scenario.(4)Design of authentication strengthening method based on model degradation and model reconstructionThis paper proposes an authentication strengthening method based on model degradation and model reconstruction.A model degradation scheme is proposed for static authentication,which solves the cold start problem to a certain extent.For continuous authentication,a model reconstruction method is proposed for the missing feature index and missing feature number of the registration set to improve the authentication accuracy and the adaptability to the registration samples.(5)Design and implementation of online testing prototype systemBased on the above methods,this paper designs and implements an online testing prototype system.The system includes a underlying platform for identity authentication based on keystroke behavior.This article elaborates the detailed design scheme of the system,displays and verifies it,and finally compares and analyzes it with other related methods.