UC Riverside UC Riverside Electronic Theses and Dissertations
Total Page:16
File Type:pdf, Size:1020Kb
UC Riverside UC Riverside Electronic Theses and Dissertations Title Towards Improving Cybersecurity and Augmenting Human Training Performance Using Brain Imaging Techniques Permalink https://escholarship.org/uc/item/1kg856mj Author Rahman, Md Lutfor Publication Date 2020 Peer reviewed|Thesis/dissertation eScholarship.org Powered by the California Digital Library University of California UNIVERSITY OF CALIFORNIA RIVERSIDE Towards Improving Cybersecurity and Augmenting Human Training Performance Using Brain Imaging Techniques A Dissertation submitted in partial satisfaction of the requirements for the degree of Doctor of Philosophy in Computer Science by Md Lutfor Rahman December 2020 Dissertation Committee: Dr. Chengyu Song, Chairperson Dr. Megan Peters Dr. Vagelis Papalexakis Dr. Srikanth Krishnamurthy Dr. Zhiyun Qian Copyright by Md Lutfor Rahman 2020 The Dissertation of Md Lutfor Rahman is approved: Committee Chairperson University of California, Riverside Acknowledgments Alhamdulillah for everything. It would be impossible for a child of parents without a formal education to attain the highest degree (Ph.D.) without the mercy of the Almighty. I am grateful to my advisor Dr. Chengyu Song - a brilliant researcher and humble being who allowed me to grow as an independent researcher, but was always by my side for any help. My Ph.D. journey was smooth and stress-free from beginning to end. I am also grateful to Dr. Megan Peters, Dr. Vagelis Papalexakis, Dr. Srikanth Krishnamurthy, Dr. Zhiyun Qian for their valuable suggestions in my research and career. It will surely help me to be a better scientist. I would like to thank my internship mentors at Army Research Lab, Dr. Antony Passaro and Dr. Benjamin T. Files, and other collaborators, Dr. Peter Khooshabehadeh, Dr. Kimberly Pollard, and Ashley Oiknine, who taught me to carefully address precise details in designing the neuroscience experiment. I want to express my gratitude to all of my formal and informal teachers in my entire educational journey. I am grateful to all our family friends in the Riverside community, with whom we spent excellent family time. I’d like to thank my co-authors, lab-mates, friends, and staff of the computer science department. I am grateful to sahadat uncle for supporting me in different ways. I like to thank all volunteers of the Education Foundation (efcharity.org) who are doing many wonderful things for underprivileged students with me. I’d like to thank my previous supervisor, Dr. Nitesh Saxena, who helped me fall in love with this challenging domain. I have to give special thanks to Dr. Ajaya Neupane, for brainstorming new ideas with me and helping me consistently throughout my Ph.D. journey. I am incredibly grateful to Dr. Mahmud Hossain for providing assistance and support in many ways. I appreciate the endorsement of Dr. Md. Mostofa Akbar, Dr. Mohammad Mahfuzul Islam, Prof. Md. Abdus Sattar, iv Dr. Ragib Hasan, and Dr. Robert M Hyatt for giving me recommendations during MS and Ph.D. admission. Throughout my educational pursuit, my family has supported me in countless ways. I want to thank my two elder brothers, Md Samsul Islam and Md Rafiqul Islam. If they had not pursued an education, I might not have had the chance to study. I am grateful to my whole family for their love and prayers. I also like to thank my relatives, and well wishers. My children, Nabiha and Rifqat, made tremendous sacrifices during my journey. Many nights I could not go to sleep with them due to work. I am so fortunate to have married my friend Umme Hani Mst. Zoaria, a superwoman whose unconditional love and tremendous support often gave me the strength to keep going one more step. v Dedication To my father, Md Zasim Uddin, and my mother, Mst. Mazeda Begum. You never had a chance for formal education but your hardship, love, and prayers have paved the way for me to attain the highest degree. vi ABSTRACT OF THE DISSERTATION Towards Improving Cybersecurity and Augmenting Human Training Performance Using Brain Imaging Techniques by Md Lutfor Rahman Doctor of Philosophy, Graduate Program in Computer Science University of California, Riverside, December 2020 Dr. Chengyu Song, Chairperson Human behaviors can weaken the security of cyber-physical systems. However, con- ventional security research focuses more on hardware and software security than analyzing and improving human behaviors to provide better protection for digital systems. In this regard, we study the neural insights of computer systems users to identify cyber-attacks, such as phishing, and improve cybersecurity. First, we analyze neural activities to detect phishing attacks. We demonstrate that the variations in neural activity levels can be utilized to identify phishing websites with improved data preprocessing and feature extraction methods. Second, we study users’ neural activities to learn their high-level intents when they use applications. The inferred intents are then used to ensure the security and privacy of sensitive resources, such as cameras and multi-media files. Finally, we design an adaptive training model that enables users to differentiate between benign and malicious scenarios. We consider both behavior and neural metrics to develop adaptive logic. Our experimental results show that participants trained with our approach outperform in the transfer task than those trained with non-adaptive and behaviorally adaptive designs. vii Contents List of Figures xi List of Tables xiii 1 Introduction1 1.1 Understanding Phishing Attacks through the Lens of BCI..............2 1.2 Neural Signals in the Loop..............................3 1.2.1 Neural Feedback for Access Control.....................4 1.2.2 Neural Feedback for Adaptive Training...................5 1.3 Thesis Contribution..................................6 1.4 Organization......................................7 2 Background8 2.1 Introduction to Brain-Computer Interface (BCI)...................8 2.1.1 Brain Imaging Techniques..........................9 2.1.2 Data Preprocessing and Feature Extraction................. 10 2.1.3 Brain Areas.................................. 11 2.2 Ethical and Safety Considerations.......................... 12 3 Learning Tensor-based Representations from Brain-Computer Interface Data for Cy- bersecurity 13 3.1 Introduction...................................... 14 3.2 Prior Work....................................... 19 3.2.1 Related Works................................ 19 3.3 Study Design & Data Collection........................... 22 3.3.1 Phishing Detection Task........................... 22 3.3.2 Study Protocol................................ 24 3.4 Behavioral Data Analysis............................... 25 3.5 Neural Data Analysis................................. 26 3.5.1 Data Preprocessing.............................. 26 3.5.2 Independent Component Analysis...................... 27 3.5.3 Feature Extraction Using Auto Regression.................. 28 viii 3.5.4 Feature Extraction Using Tensor Decomposition.............. 29 3.6 Results and Analysis................................. 35 3.6.1 Classifiers and Evaluation Metrics...................... 35 3.6.2 Global Model................................. 37 3.6.3 Human vs Machine.............................. 37 3.6.4 Comparison of Auto-regression and Tensor Decomposition Result..... 39 3.7 Discussion....................................... 41 3.7.1 Phishing Detection Mechanism....................... 41 3.7.2 Feasibility of the Defense Mechanism.................... 42 3.7.3 Phishing Detection vs Brain Areas...................... 42 3.7.4 Statistical Analysis: Real vs Fake Events.................. 44 3.7.5 Feature Space Reduction........................... 45 3.7.6 Study Strengths and Limitations....................... 45 3.8 Chapter Conclusion.................................. 46 4 IAC: On the Feasibility of Utilizing Neural Signals for Access Control 48 4.1 Introduction...................................... 49 4.2 Background...................................... 53 4.3 Intent-driven Access Control............................. 55 4.3.1 Threat Model and Assumptions....................... 56 4.3.2 IAC via BCI................................. 57 4.4 Experiment Design.................................. 59 4.4.1 Single App Experiment............................ 60 4.4.2 Multiple Apps Experiment.......................... 62 4.4.3 Experimental Procedures........................... 63 4.5 Data Process and Analysis.............................. 65 4.6 Feasibility Test.................................... 68 4.6.1 Single App Analysis............................. 69 4.6.2 Cross-app Portability Analysis........................ 70 4.6.3 Results Analysis............................... 72 4.6.4 Authorization Accuracy........................... 73 4.7 Discussion....................................... 73 4.8 Related Work..................................... 74 4.9 Chapter Conclusion.................................. 77 5 Augmenting Training Performance by Adding Neural Signals into the Adaptive Feed- back Loop 79 5.1 Introduction...................................... 80 5.2 Background...................................... 83 5.2.1 Training.................................... 83 5.2.2 Closed-loop BCI............................... 85 5.2.3 Theta/Alpha Ratio (TAR)........................... 85 5.3 Design of the Experiments.............................. 86 5.3.1 Go/No-Go Training Task........................... 86 5.3.2 Target Identification Transfer Task...................... 90 ix 5.4 Methods.......................................