Evaluating the Use of Autobiographical Memory for Authentication

This paper takes a first step in exploring what we call autobiographical authentication, using data about events and activities that are easily captured by smartphones and other devices as the basis for authentication. More specifically, this paper reports on two studies. The first study looked at understanding and categorizing what kinds of everyday memories people could recall, using the results to generate an ontology and understand the relationship between various factors and memorability. The second study investigated the difficulty level in answering different kinds of questions about everyday memories, along with the types of answers that can be expected and the level of confidence that can be ascribed to these answers. We found that there are 9 categories that comprise what users recall about their everyday memories that can be easily “captured” by digital sensors, that meta-knowledge of the response and participant demographics can affect memorability, that users can respond to autobiographical questions about themselves with fairly high confidence, and that users tend to respond to everyday memory questions with events in the past day or so.