Research On Identity Authentication Method Based On Keystroke Behavior

Nowadays,while protecting sensitive data and computer systems from imposters,improving the identity authentication methods easy to use has also become an important challenge in the field of computer and information security.Research shows that the use of keystroke dynamics for identity authentication can not only ensure security,but also has high practicability.Keystroke dynamics uses human keystroke rhythm to authenticate identity.It focuses not on what the user types,but how the user types.In this method,the user only needs to enter the text as usual,without other additional operations and feature extraction equipment.Keystroke dynamics authentication is generally divided into fixed text authentication and free text authentication.Compared with fixed text authentication,free text authentication can handle more free content and time periods,so it can be applied to more scenarios,such as continuous authentication systems.Nowadays,the research on keystroke dynamics authentication has gradually changed from fixed text authentication to free text authentication.However,the features in free text are difficult to collect and train,and the length of the collected text is difficult to weigh.For these problems,this thesis has carried on the related research,the specific research content is as follows:1.A multi decisions continuous identity authentication method based on keystroke behavior is proposed.Firstly,the key values entered by users are assigned different labels by K-means clustering method,and labels are used to participate in the creation and training of user templates,so as to solve the problem of feature collection,and then the user template is created and verified by single classification algorithm.Finally,the idea of three-way decisions is introduced,and the calculation formula of the threshold is given,so as to achieve the balance between authentication accuracy and system security.2.A further processing method of keystroke behavior characteristics is proposed.In view of the great difference in the number of keys in different clusters when the original K-means clustering algorithm was used to deal with keystroke characteristics,this thesis made an improvement.Considering that the difference between some key values is small,and they belong to one cluster,they are also very close to other different clusters.Therefore,the fuzzy clustering algorithm is introduced.The same key value can belong to multiple clusters at the same time,so as to reduce the gap between the number of key values in different clusters.Finally,more key-pairs can meet the training conditions.Because more tag combinations are trained in the system,the types of input key-pairs that can be processed are increased to a certain extent,which improves the security of the system on the other hand.