Application of deep learning to user authentication systems using s-EMG

2020 Fiscal Year Final Research Report

• Project/Area Number: 17K00186
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Information security
• Research Institution: University of Miyazaki
• Principal Investigator: Yamaba Hisaaki 宮崎大学, 工学部, 助教 (60260741)
• Co-Investigator(Kenkyū-buntansha): 岡崎 直宣 宮崎大学, 工学部, 教授 (90347047) ; 油田 健太郎 宮崎大学, 工学部, 准教授 (30433410)
• Project Period (FY): 2017-04-01 – 2021-03-31
• Keywords: バイオメトリクス / 深層学習 / 筋電位 / 個人認証 / 携帯端末

Outline of Final Research Achievements

To prevent shoulder-surfing attacks, we proposed a user authentication method using surface electromyogram (s-EMG) signals, which can be used to identify who generated the signals and which gestures were made. Our method uses a technique called ‘pass-gesture’, which refers to a series of hand gestures, to achieve s-EMG-based authentication. However, it is necessary to introduce computer programs that can recognize gestures from the s-EMG signals. In this study, we propose two methods that can be used to compare s-EMG signals and determine whether they were made by the same gesture. One uses support vector machines (SVMs), and the other uses dynamic time warping. We also introduced an appropriate method for selecting the validation data used to train SVMs using correlation coefficients and cross-correlation functions. Deep learning method, which is expected to detect suitable feature values, was also introduced to improve the performance of gesture identification.

Free Research Field

コンピュータセキュリティ

Academic Significance and Societal Importance of the Research Achievements

本研究では、パスワードに替わる認証情報として、前腕部の動作の組み合わせである「パスジェスチャ」の導入を行なった。これにより、覗き見によるパスワードの漏洩に対抗しうる、バイオメトリクスの一つである筋電位を用いた「安全なユーザ認証」が期待できる。さらにこのアプローチの長所は、単に筋電波形の個人差 (Something you are) だけに依存して認証を行うのではなく、ジェスチャをどの順序で組み合わせてパスワードとするのか (Something you know) を認証に活用することにより、多要素認証の実現の可能性を示すことができた。