A new statistical anomaly detector model for keystroke dynamics on touch mobile devices
Other Title(s)
نموذج جديد لكاشف تباين إحصائي لديناميكية الكتابة باللمس على الهواتف النقالة
Dissertant
Thesis advisor
Comitee Members
Shkukani, Muhammad
al-Hammuz, Sadiq O.
University
Middle East University
Faculty
Faculty of Information Technology
Department
Computer Science Department
University Country
Jordan
Degree
Master
Degree Date
2016
English Abstract
Keystroke Dynamics – the authentication technology that utilizes the typing rhythm to distinguish genuine users from impostors, has gone through continued developments to improve its detection capability.
Recently, the keystroke dynamics model has been investigated as an authentication method on touch mobile devices, which resulted in shifting the attention from enhancing classifiers only, to adding new measurable features of mobile devices that can improve the classifiers’ detection performance.
The work in this thesis investigates keystroke dynamics, through empirical analysis of experimental datasets collected on mobile devices which included timing features as well as key-press pressure and finger area.
A statistical median-based binary classifier (anomaly detector) is proposed, the Med-Min-Model, which utilizes the distance to the median in calculating the upper and lower thresholds of a feature.
The two thresholds are determined in the training phase, and used later in the authentication (testing) phase to classify feature values that result from typing during the testing phase, as genuine or impostor.
An existing dataset is utilized in evaluating the Equal-Error-Rate (EER) of the proposed model in comparison with three verification models.
The resulting EER value of the proposed model, using the existing dataset is 0.0679, which is much lower than EER value of the three verification models.
The proposed model is implemented as a data collection and authentication system, for use on a touch tablet working under the Andriod operating system, which measured typing timing features, pressure, and finger area.
The system is used in the collection of a new dataset (MEU-Mobile) from 56 subjects where each subject typed on the tablet a unified password 51 times (34 training attempts and 17 testing attempts).
Analysis of the new dataset shows a reduced EER value of 0.0494 compared to the EER value using the existing dataset The False-Acceptance-Rate (FAR) at 5% False-Rejection-Rate (FRR) was 5.79%, which points to the fact that further enhancement is needed to reduce the False-Acceptance-Rate.
The proposed model used a pass-mark as a reference value for the resulting test-score of a typing attempt.
Two methods were used in determining the pass-mark; a variable pass-mark for each subject which is tuned to get to the point of equal FAR and FRR, and a global (fixed) pass-mark for all subjects, that is derived from the average of pass-marks of all subjects.
An analysis using a global pass-mark showed a slightly higher EER (0.0548).
The thesis ends with presenting conclusions and recommendations for future work based on results of the present research
Main Subjects
Information Technology and Computer Science
No. of Pages
88
Table of Contents
Table of contents.
Abstract.
Abstract in Arabic.
Chapter One : Introduction.
Chapter Two : Background and literature review.
Chapter Three : The proposed keystroke dynamics model.
Chapter Four : Experimental results and discussion.
Chapter Five : Conclusion and future work.
References.
American Psychological Association (APA)
al-Ubaydi, Nur Mahmud Shakir. (2016). A new statistical anomaly detector model for keystroke dynamics on touch mobile devices. (Master's theses Theses and Dissertations Master). Middle East University, Jordan
https://search.emarefa.net/detail/BIM-721103
Modern Language Association (MLA)
al-Ubaydi, Nur Mahmud Shakir. A new statistical anomaly detector model for keystroke dynamics on touch mobile devices. (Master's theses Theses and Dissertations Master). Middle East University. (2016).
https://search.emarefa.net/detail/BIM-721103
American Medical Association (AMA)
al-Ubaydi, Nur Mahmud Shakir. (2016). A new statistical anomaly detector model for keystroke dynamics on touch mobile devices. (Master's theses Theses and Dissertations Master). Middle East University, Jordan
https://search.emarefa.net/detail/BIM-721103
Language
English
Data Type
Arab Theses
Record ID
BIM-721103