Font Size: a A A

Dynamic Authentication Based On Keystroke And Mouse Dynamics

Posted on:2019-09-05Degree:MasterType:Thesis
Country:ChinaCandidate:G ChenFull Text:PDF
GTID:2428330590492391Subject:Electronic and communication engineering
Abstract/Summary:PDF Full Text Request
The widespread application of information systems not only brings convince to our life but great challenges to information security.Authentication is of great importance as it serves as the first line of defense.However,traditional static authentication methods mainly rely on the username and password scheme,which suffers from the risks of being stolen or cracked.Moreover,static authentication cannot detect the malicious behaviors occurring after the log-in time.In contrast,dynamic authentication is one technique that re-authenticates the user continuously throughout the lifetime of the user's activity,which can serve as an auxiliary approach for conventional static mechanisms.Authentication based on behavioral biometrics aims to identify users by analyzing the patterns of their activities,which is appropriate for the implementation of dynamics authentication.This paper proposes a dynamic authentication system based on both keystroke and mouse dynamics,which identifies users based on the analysis of their operations with a keyboard and mouse.The proposed authentication system mainly consists of two verification modules.The module based on keystroke dynamics extracts keystroke features from users' typing patterns,utilizes KPCA(Kernel Principal Component Analysis)to reduce the dimension of the keystroke feature space,and employs LS-SVM(Least Squares-Support Vector Machine)as the classifier.Within the module based on mouse dynamics,weighted ELM(Extreme Learning Machine)is conducted as the classifier to deal with the imbalanced behavioral data.A score fusion strategy is applied to combine the output from two modules for authentication.Preliminary experiments are conducted to evaluate the performance of the proposed system and an average verification accuracy of 99.76% is achieved.
Keywords/Search Tags:authentication, behavioral biometrics, keystroke, mouse dynamics
PDF Full Text Request
Related items