作者: Paolo Gasti , Jaroslav Sedenka , Zdenka Sitova , Ge Peng , Qing Yang
DOI:
关键词: Embedded system 、 Computer science 、 Biometrics 、 Modality (human–computer interaction) 、 Computer vision 、 Key generation 、 Key (cryptography) 、 Keystroke logging 、 Authentication 、 Artificial intelligence 、 Feature extraction
摘要: In this paper, we introduce Hand Movement, Orientation, and Grasp (HMOG), a behavioral biometric to continuously authenticate smartphone users. HMOG unobtrusively captures subtle micro-movement orientation dynamics resulting from how user grasps, holds, taps on the smartphone. We evaluated authentication key generation (BKG) performance of features typing data collected 100 subjects. Data was under two conditions: sitting walking. achieved EERs as low 6.92% (walking) 10.20% (sitting) when combined HMOG, tap, keystroke features. performed additional experiments investigate why perform well during Our results suggest that is due ability capture distinctive body movements caused by walking, in addition hand-movement taps. With BKG, 13.7% using with comparison, BKG hold, swipe led between 31.3% 36.7%. also analyzed energy consumption feature extraction computation. analysis shows extracted at 16Hz sensor sampling rate incurred minor overhead 7.9% without sacrificing accuracy.