Date of Award
Computer and Information Science
Security systems for modern computing devices suffer from a multitude of weaknesses that can render users helpless against an attack on their system. Various attempts at incorporating human characteristics into security systems have achieved varying levels of success in improving security. In this paper, we study the usefulness of TouchAnalytics™ - a second-level security system that attempts to authenticate a user based on touch-data gathered from an interaction with the device. Through the use of machine learning algorithms, we developed a system that is successful at au- thenticating users, achieving under 0.05% False Authentication Rate (FAR). We conclude that the removal of strictly defined training activities yields more characteristic data for each user, and thus, a more accurate system.
McLeod, Clay, "Dynamic Feature Extraction and User Classification using TouchAnalytics™" (2014). Honors Theses. 1253.