Good automatic authentication question generation

Abstract

We explore a novel application of Question Generation (QG) for authentication use, where questions are widely used to verify user identity for online accounts. In our approach, we prompt users to provide a few sentences about their personal life events. We transform user-provided input sentences into a set of simple fact-based authentication questions. We compared our approach with previous QG systems, and evaluation results show that our approach yielded better performance and the promise of future personalized authentication question generation.