Table 1 Overview of studies on swipe gesture recognition on mobile devices
From: Evaluation of stability of swipe gesture authentication across usage scenarios of mobile device
Publication | Year | Number of subjects | Data collection mode | Scenarios | Feature extraction | Classification | EER |
|---|---|---|---|---|---|---|---|
Frank et al. [6] | 2013 | 41 | Constrained data collection | Not specified | Timestamp, X, Y, phone orientation | SVM RBF kernel and k-NN | 0%-Intra-session and 2–3%-inter-session % |
Bo et al. [7] | 2014 | 100 | Constrained data collection | Static (no body movement) and walking | Timestamp, X, Y, finger pressure | SVM | Static scenario (FAR -Tap 22, Fling-9, Scroll- 23), walking scenario accuracy 100% after 12 steps of walking |
Saravanan et al. [8] | 2014 | 20 | Constrained | Not specified | Timestamp, X, Y, pressure | SVM, random forests and BayesNet | 97.9% accuracy mobile phones 96.79% - tablets |
Feng at al. [9] | 2014 | 23 phone owners, 100 guest users | Unconstrained | Real-life scenarios | Timestamp, X, Y, size, pressure, swipe length, swipe curvature | DTW with one nearest neighbour | 90% accuracy |
Wang et al. [10] | 2017 | 160 set of app usage data | Unconstrained | Not Specified | Timestamp, X, Y, phone orientation | SVM RBF Kernel | AUC (area under the curve) score of 80% to 96% (detecting unauthorised access) % |
Our model | 2018 | 50 | Unconstrained | Sitting, walking, treadmill, travelling on a bus | Timestamp, X, Y, finger area, finger pressure | SVM, kNN and naive Bayes | 1 % |