TABLE 10.
Person authentication (PA).
| Reference | AI Algorithm | Dataset/Data Modality | Purpose |
|---|---|---|---|
| Zhang et al. (2020b) | SVDD and PCA for illegal user detection, LSTM for PI | Velocity and acceleration from the smartphone at the leg | PI and illegal user detection |
| Li et al. (2020) | Two-stream CNN with SVM | BrainRun dataaset. Own dataset of gait and other behavioral features from smartphones, 100 subjects. | SCANet: Continuous PA, distinguishes legitimate vs impostor users |
| (Zhang et al., 2014b; Qin et al., 2019) | Multi-layer LSTM and Extreme Value Statistic | ZJU-GaitAcc, 3D accelerations from smartphones | PI and PA of the learned user, reject unauthorized user |
| (Wu et al., 2018; Hintze et al., 2019) | SVM, KNN, DT | acceleration, angular velocity, magnetic intensity, and PPG signals from fingertip device | Multisensor PA, HAR |
| Vandersmissen et al. (2018) | Deep CNN | Own IDRad Dataset: micro-Doppler signatures, 5 subjects | Automatic intruder detection, indoor PI |
| Jorquera Valero et al. (2018) | Semi-supervised ML, Isolation Forest | Tracking current vs. known usage of the device and motion sensor data from phone | Adaptive and continuous PA system, anomaly detection |
| Neverova et al. (2016) | Dense clockwork RNN | HMOG, Google Abacus Dataset: time series of inertial measurements | distinguishes legitimate vs impostor users |
Legend: Photoplethysmography (PPG), Support Vector Data Description (SVDD), Growing Neural Gas (GNG).
Datasets: BrainRun (Papamichail et al., 2019), ZJU-GaitAcc (Zhang Y. et al., 2014), HMOG (Yang et al., 2014), UMN (Raghavendra et al., 2006), UCSD Ped (Li et al., 2013), Avenue (Lu et al., 2013).