Date Awarded

2017

Document Type

Dissertation

Degree Name

Doctor of Philosophy (Ph.D.)

Department

Computer Science

Advisor

Haining Wang

Abstract

With the expanding popularity of various Internet services, online users have be- come more vulnerable to malicious attacks as more of their private information is accessible on the Internet. The primary defense protecting private information is user authentication, which currently relies on less than ideal methods such as text passwords and PIN numbers. Alternative methods such as graphical passwords and behavioral biometrics have been proposed, but with too many limitations to replace current methods. However, with enhancements to overcome these limitations and harden existing methods, alternative authentications may become viable for future use. This dissertation aims to enhance the viability of alternative authentication systems. In particular, our research focuses on graphical passwords, biometrics that depend, directly or indirectly, on anthropometric data, and user authentication en- hancements using touch screen features on mobile devices. In the study of graphical passwords, we develop a new cued-recall graphical pass- word system called GridMap by exploring (1) the use of grids with variable input entered through the keyboard, and (2) the use of maps as background images. As a result, GridMap is able to achieve high key space and resistance to shoulder surfing attacks. To validate the efficacy of GridMap in practice, we conduct a user study with 50 participants. Our experimental results show that GridMap works well in domains in which a user logs in on a regular basis, and provides a memorability benefit if the chosen map has a personal significance to the user. In the study of anthropometric based biometrics through the use of mouse dy- namics, we present a method for choosing metrics based on empirical evidence of natural difference in the genders. In particular, we develop a novel gender classifi- cation model and evaluate the model’s accuracy based on the data collected from a group of 94 users. Temporal, spatial, and accuracy metrics are recorded from kine- matic and spatial analyses of 256 mouse movements performed by each user. The effectiveness of our model is validated through the use of binary logistic regressions. Finally, we propose enhanced authentication schemes through redesigned input, along with the use of anthropometric biometrics on mobile devices. We design a novel scheme called Triple Touch PIN (TTP) that improves traditional PIN number based authentication with highly enlarged keyspace. We evaluate TTP on a group of 25 participants. Our evaluation results show that TTP is robust against dictio- nary attacks and achieves usability at acceptable levels for users. We also assess anthropometric based biometrics by attempting to differentiate user fingers through the readings of the sensors in the touch screen. We validate the viability of this biometric approach on 33 users, and observe that it is feasible for distinguishing the fingers with the largest anthropometric differences, the thumb and pinkie fingers.

DOI

http://dx.doi.org/doi:10.21220/S26M2W

Rights

© The Author

Share

COinS