Font Size: a A A

Research And Implementation Of Identity Authentication Technology Based On Keystroke Dynamics

Posted on:2019-07-08Degree:MasterType:Thesis
Country:ChinaCandidate:W H HuangFull Text:PDF
GTID:2348330542998706Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
The traditional identity-based authentication technology based on keystroke behavior is mainly based on temporal features,such as obtaining the press time of a button or the time interval between different buttons by obtaining the press time and release time of different buttons,and the feature types acquired in this way is too simple,and time characteristics are more susceptible to the user's own and external factors,the stability is poor,the user behavior pattern recognition based on it is less effective.In view of the above problems,this paper improves and innovates the design of feature selection and classification algorithms,enriches the types of extractable features,enhances the stability of the selected features,and improves the impact of the relatively stable features on the classification results.Thereby,the identification effect of the authentication technology is optimized and the detection accuracy is improved.At the same time,in order to improve the practicability of the designed identity authentication technology,this paper chooses the free-text-based keystroke dynamics authentication method.The research work and achievements made in this paper are as follows:In the aspect of feature extraction of keystroke behavior,in terms of keystroke feature extraction,this paper further expands and optimizes the combination of non-traditional features and proposes the feature set used in this paper.Compared with temporal features,this feature set enriches the types of features extracted.At the same time,the non-traditional features included in this feature set are extracted based on the user's long text input,so it has good stability and can reflect the user's behavior characteristics more objectively.Through data analysis and corresponding comparison experiments,it can be seen-that the-non-traditional feature-data sets extracted in this paper have improved the recognition accuracy of user identities to a certain extent under the test of SVM and DT classification algorithms.In the aspect of design of authentication technology classification algorithm,the algorithm used in this paper is designed based on a single classification model.By computing the minimum weighted distance between the samples to be tested and the training samples,then compares it with the preset threshold to determine the user's identity.The so-called weighted distance,that is,the original distance calculated of each dimension is weighted by dividing standard deviation of the dimension's features.The algorithm has three advantages:Firstly,the algorithm is designed based on a single classification model,which is more compatible with the actual use scenarios and less difficult to collect training data.Second,the classification algorithm calculates the minimum distance between the user to be detected and the training data then it can better identify the users.Finally,it is also the key point of the algorithm.The algorithm weighted the distance calculation,which can highlight the influence of the relatively stable features on the distance calculation results,so as to enhance its effect on the final classification results and give better play to the data advantage of the relatively stable features.To a certain extent,the instability problem of the characteristic data is further solved.Experimental results show that the classification algorithm designed in this paper has reduced the EER rate to 10.24%,compared to other research achievements in this field,it has been improved to some extent.Combined with the non-traditional features and the single-classification algorithm based on the minimum weighted distance,this paper designs and implements the identity authentication system based on the keystroke dynamics,and effectively combines the research work done in this paper,the test results show that the system can identify the user's identity in 410ms,and the FAR and FRR rates were reduced to 10.75%and 7.26%,the system runs up to 96.5M of maximum memory usage,with good practicality,provides a certain reference for popularization and application of keystroke dynamics authentication technology.
Keywords/Search Tags:Keystroke identification, Free text detection, Nontraditional feature, Minimum weighted distance
PDF Full Text Request
Related items