Authentication by Touch Operation on Smartphone with Support Vector Machine
Nozomi Miyamoto, Chihiro Shibata, Toshiyuki Kinoshita · International Journal for Information Security Research · 2017
In this research, we proposed a method to apply Support Vector Machine (SVM) for personal authentication by touch operation on smartphone.The SVM is one of machine learning that generates a maximum margin separation line furthest from both of the separated sample point clusters.The data is divided into training data and test data, and the authentication accuracy is measured by applying Cross Validation.For four kinds of touch operations: single tap, double tap, swipe, rotation, and six pairs of these touch operations, we compared the authentication rates of each touch operation and pair of operations and measured how the authentication rate changes along with the number of training data and the number of test data.The experimental results show that the authentication rate is as low as 60 to 85% for a single touch operation, but when the pair of touch operations are applied, the authentication rate reaches 95 to 100% at the maximum.It can be used sufficiently for personal authentication.As the number of registered data and test data increases, the authentication rate also increases.However, considering user's burden at registration and the registration time, it is considered that the combination of the double tap and swipe, 20 registered data and 2 test data are optimal in practical use.