Cost-Effective Modeling for Authentication and Its Application to Activity Tracker

Abstract

The growing popularity of multi-factor authentication, it makes the need for more cost-effective approach. The more authentication factors the system use, the higher cost for machine learning all of factors it require. We considered the behavioral authentication that is brought to attention as one of the authentication factors. The behavioral authentication works well with machine learning approaches. However, with machine learning, a model must be created, and the service provider must analyze each user individually; both adding to the cost. In this paper, we propose a cost-effective user modeling approach that uses a FuelBand to obtain activity information for behavioral authentication. This approach uses a clustering method that focuses on the characteristics of our behavioral authentication method. The performance of our system was compared to that of machine learning 70 users, and for no more than half the cost, the results had an accuracy of 89.28i¾?%.