Mobile Sensor-Based Biometrics from Common Daily Activities

Kenichi Yoneda, Fordham University


Research on mobile sensor biometrics increased when mobile devices with powerful sensors, such as smartphones, became ubiquitous. However, existing studies are quite limited, especially with regard to the physical activities that are used to provide the biometric signature—many studies only consider a single activity. In this study, we provide the most comprehensive study of mobile biometrics to date. We evaluate eighteen physical activities and nine sensor combinations for their biometric efficacy (the accelerometer and gyroscope sensors from a smartphone and smartwatch are used). Our mobile biometric models are evaluated with respect to identification and authentication performance and are shown to achieve excellent results in both cases. Furthermore, our models perform well even when built from all eighteen activities without activity labels, which represents a big step towards achieving the goal of continuous biometrics using only a smartwatch and smartphone.

Subject Area

Computer science

Recommended Citation

Yoneda, Kenichi, "Mobile Sensor-Based Biometrics from Common Daily Activities" (2017). ETD Collection for Fordham University. AAI10278712.