B-2-2 スマートフォンにおけるリズム認証手法実用化のための検討(自己組織化マップ(1))

抄録

In this study, we have developed and proposed a secure and useful authentication method to resist 'shoulder surfing' attacks for smartphone. In our previous works, we adopted Self-Organizing Map for an analytical technique and the precision of the authentications has been sufficient to ordinary use. As a step toward practical use in near future, it will be necessary to compare our technique with the other orthodox analytic methods at first. In this article, we compare the k-means with the SOM method and we evaluate both the precision of the 2-dimensional Euclidean distance, which is derived by SOM, and another precision of the 7-dimensional Euclidean distance derived directly. As a result, the precision of SOM method was 98.3% whereas the precision of 7-dimensional direct method was 95.6%. It was confirmed that the k-means method was insufficient for clustering relative to the SOM method at the evaluation experiment, and it was suggested that the SOM method was more suitable for the program implementation.