(EEG) Based Biometrics Investigation for Authentication

228 Pages·2016·8.5 MB·English

Nova Southeastern University
NSUWorks
CEC Theses and Dissertations College of Engineering and Computing
2015
An Electroencephalogram (EEG) Based
Biometrics Investigation for Authentication: A
Human-Computer Interaction (HCI) Approach
Ricardo J. Rodriguez
Nova Southeastern University,[email protected]
This document is a product of extensive research conducted at the Nova Southeastern UniversityCollege of
Engineering and Computing. For more information on research and degree programs at the NSU College of
Engineering and Computing, please clickhere.
Follow this and additional works at:http://nsuworks.nova.edu/gscis_etd
Part of theInformation Security Commons
Share Feedback About This Item
NSUWorks Citation
Ricardo J. Rodriguez. 2015.An Electroencephalogram (EEG) Based Biometrics Investigation for Authentication: A Human-Computer
Interaction (HCI) Approach.Doctoral dissertation. Nova Southeastern University. Retrieved from NSUWorks, College of Engineering
and Computing. (67)
http://nsuworks.nova.edu/gscis_etd/67.
This Dissertation is brought to you by the College of Engineering and Computing at NSUWorks. It has been accepted for inclusion in CEC Theses and
Dissertations by an authorized administrator of NSUWorks. For more information, please [email protected].
An Electroencephalogram (EEG) Based Biometrics Investigation for Authentication: A
Human-Computer Interaction (HCI) Approach
by
Ricardo J. Rodriguez
A dissertation submitted in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
In
Computer Information Systems
College of Engineering and Computing
Nova Southeastern University
2015
We hereby certify that this dissertation, submitted by Ricardo Rodriguez, conforms to acceptable
standards and is fully adequate in scope and quality to fulfill the dissertation requirements for
the degree of Doctor of Philosophy.
_____________________________________________ ________________
Maxine S. Cohen, Ph.D. Date
Chairperson of Dissertation Committee
_____________________________________________ ________________
Sumitra Mukherjee, Ph.D. Date
Dissertation Committee Member
_____________________________________________ ________________
Bruce Montgomery, Ph.D. Date
Dissertation Committee Member
Approved:
_____________________________________________ ________________
Amon B. Seagull, Ph.D. Date
Interim Dean, College of Engineering and Computing
College of Engineering and Computing
Nova Southeastern University
2015
An Abstract of a Dissertation Submitted to Nova Southeastern University in Partial
Fulfillment of the Requirements for the degree of Doctor of Philosophy in Computer
Information Systems
An Electroencephalogram (EEG) Based Biometrics Investigation for
Authentication: A Human-Computer Interaction (HCI) approach
by
Ricardo J. Rodriguez
August 2015
Encephalogram (EEG) devices are one of the active research areas in human-
computer interaction (HCI). They provide a unique brain-machine interface (BMI) for
interacting with a growing number of applications. EEG devices interface with
computational systems, including traditional desktop computers and more recently
mobile devices. These computational systems can be targeted by malicious users. There
is clearly an opportunity to leverage EEG capabilities for increasing the efficiency of
access control mechanisms, which are the first line of defense in any computational
system.
Access control mechanisms rely on a number of authenticators, including “what
you know”, “what you have”, and “what you are”. The “what you are” authenticator,
formally known as a biometrics authenticator, is increasingly gaining acceptance. It uses
an individual’s unique features such as fingerprints and facial images to properly
authenticate users. An emerging approach in physiological biometrics is cognitive
biometrics, which measures brain’s response to stimuli. These stimuli can be measured
by a number of devices, including EEG systems.
This work shows an approach to authenticate users interacting with their
computational devices through the use of EEG devices. The results demonstrate the
feasibility of using a unique hard-to-forge trait as an absolute biometrics authenticator by
exploiting the signals generated by different areas of the brain when exposed to visual
stimuli. The outcome of this research highlights the importance of the prefrontal cortex
and temporal lobes to capture unique responses to images that trigger emotional
responses.
Additionally, the utilization of logarithmic band power processing combined with
LDA as the machine learning algorithm provides higher accuracy when compared against
common spatial patterns or windowed means processing in combination with GMM and
SVM machine learning algorithms. These results continue to validate the value of
logarithmic band power processing and LDA when applied to oscillatory processes.
Acknowledgements
When I decided to pursue a Ph.D. with a concentration in Information Assurance,
I knew it was going to be a challenging task, primarily due to the fact that I decided to do
so while continuing my full-time career as a cyber security engineer. My success in this
endeavor is the culmination of years of efforts and sacrifices, and required the help and
support of many people.
I would like to first and foremost thank my advisor Dr. Maxine Cohen. Her
guidance and knowledge combined with her unique set of interpersonal skills have left a
lasting impression on me. I consider her a role model to emulate. I would also like to
thank Dr. Sumitra Mukherjee and Dr. Bruce Montgomery. As dissertation committee
members, they provided invaluable input that ultimately enhanced the execution of this
work.
On the personal side, I would like to thank Dr. Arturo Ponce. He has been a close
friend for the last 20 years. Having someone going through similar experiences along the
way truly made a difference. I would also like to thank my parents, who instilled in me
the desire to continue learning and challenging myself as an essential element of my life
journey. There is also another special person who is no longer with us, Moraima Rosado.
As an extraordinary human being, friend, and fellow engineer, I dedicate this work to
you. Lastly, the two most important people in my life, my wife Xiomi and daughter
Andrea. Thank you so much for your patience, selflessness, and full support. It is for you
that I keep improving myself with the full knowledge that without your backing, it would
not be possible.
Table of Contents
Abstract iii
Acknowledgments iv
List of Tables vii
List of Figures x
Chapters
1. Introduction 1
Problem Statement 2
Dissertation Goal 3
Research Questions 4
Relevance and Significance 8
Barriers and Issues 10
Limitations and Delimitations 11
Limitations 11
Delimitations 12
Definition of Terms 12
Summary 13
2. Review of the Literature 15
Access Control 15
EEG Devices in Research 17
Biometrics Authentication 21
EEG Based Biometrics 23
Biometrics Testing and Evaluation 31
Support Vector Machines 35
Summary 38
3. Methodology 40
Modeling 40
Implementation 46
Testing 50
Analysis 54
Resources 58
Summary 59
4. Results 61
Overview 61
Data Collection Process Overview 61
Demographics 63
Biometrics Data Pre-processing 65
Feature Extraction and Classification 67
Summary 97
v
5. Conclusions, Implications, Recommendations, and Summary 99
Conclusions 99
Implications 103
Recommendations 104
Summary 105
Appendices
A. BCILAB Features 109
B. Test Execution Flowcharts 115
C. Invitation to Participate in Study 117
D. Demographics Questionnaire 118
E. Technology Acceptance Model Questionnaire 119
F. Snodgrass and Vaderwart Picture Set 147
G. User’s Perception Questionnaire 156
H. Adult/General Informed Consent 159
I. Research Protocol Approval Letter from Nova Southeastern University 162
J. Test Facility Authorization Letter 163
K. Log-Bandpower Processing Results 164
L. Common Spatial Patterns (CSP) Processing Results 167
M. Windowed Means Processing Results 170
N. RAW EEG and Best Result Model Data Visualization per Participant 173
O. Participant Answers to the Technology Acceptance Model Questionnaire 193
References 210
vi
List of Tables

Tables
1. Summary Table of Common Physiological Biometrics Used in the Research and
Development of Authentication Systems 22

2. Summary Table of Common Behavioral Biometrics Used in the Research and
Development of Authentication Systems 23

3. Summary Table of Verification Results of Different Fusion Methods for the
XM2VTS Database 25

4. Summary Table of Key Studies performed on Applicable Research Areas 37

5. BCILAB paradigms used in the study 42

6. Attempted filters using BCILAB 43

7. Attempted machine learning related functions in BCILAB 44

8. Participants per Age and Gender 64

9. Participants per Level of Education 64

10. Participants per Major and Gender 65

11. BCILAB Paradigms Used in the Study 68

12. Attempted filters using BCILAB 68

13. Attempted Machine Learning Related Functions in BCILAB 69

14. BCILAB Best Results with Log-Bandpower (ParadigmBandpower) 74

15. BCILAB Best Results with Common Spatial Patterns (ParadigmCSP) 75

16. BCILAB Best Results with Windowed Means (ParadigmWindowmeans) 76

vii
17.BCILAB Paradigm Accuracy 77
18.EER Results 88
19.Classifier Accuracy per Paradigm 89
20.Perceived Usefulness 90
21.Perceived Ease of Use 91
22.Perceived Need for Security 92
23.Perceived Need for Privacy 93
24.Perceived Invasiveness 95
25.User’s Perception Questionnaire Results – Part 1 95
26.User’s Perception Questionnaire Results – Part 2 96
27.Perception Differences between Males and Females 97
A1. EEG Raw Signal Processing Approaches Available in BCILAB 109
A2. Available Filters in BCILAB 111
A3. Available Machine Learning Related Functions in BCILAB 112
K1. BCILAB Results Using Log-Bandpower Paradigm and LDA as the Machine
Learning Function 164
K2. BCILAB Results Using Log-Bandpower Paradigm and SVM as the Machine
Learning Function 165
K3. BCILAB Results Using Log-Bandpower Paradigm and GMM as the Machine
Learning Function 166
L1. BCILAB Results Using CSP Paradigm and LDA as the Machine Learning
Function 167
viii
L2. BCILAB Results Using CSP Paradigm and SVM as the Machine Learning
Function 168
L3. BCILAB Results Using CSP Paradigm and GMM as the Machine Learning
Function 169
M1. BCILAB Results using Windowed Means Paradigm and LDA as the Machine
Learning Function 170
M2. BCILAB Results Using Windowed Means Paradigm and SVM as the Machine
Learning Function 171
M3. BCILAB Results Using Windowed Means Paradigm and GMM as the Machine
Learning Function 172
O1. Answers to Questions Related to Perceived Usefulness 193
O2. Answers to Questions Related to Perceived Ease of Use 195
O3. Answers to Questions Related to Perceived Need for Security 197
O4. Answers to Questions Related to Perceived Need for Privacy 199
O5. Answers to Questions Related to Perceived Invasiveness 201
O6. Answers to Questions Related to User’s Perception Questionnaire - Part 1 207
O7. Answers to Questions Related to User’s Perception Questionnaire - Part 2 208
ix