Understanding Keystroke Dynamics for Smartphone Users Authentication and Keystroke Dynamics on Smartphones Built-In Motion Sensors

Table 1

Comparison of studies for keystroke dynamics. Motion data column indicates whether features from motion data are used or not.

Authors	Year	Methodology	Motion data	Number of subjects	Number of training samples	Classifier	EER (%)
Clarke et al. [8]	2003	4-digit PIN	X	30	30	Statistical	11.3
Clarke and Furnell [9]	2007	4-digit PIN	X	30	30	Neural network	8.5
		6 alphabetic characters				Neural network	15.2
Chang et al. [6]	2012	3–6 thumbnails	X	100	5	Statistical	6.9
De Mendizabal- Vázquez et al. [10]	2014	4-digit PIN	O	80	3–9	Euclidean distance	20
Zheng et al. [5]	2014	4-digit PIN/; 8-digit PIN	O	80	80	Nearest neighbor distance	3.65/; 4.45
Samura et al. [7]	2014	300 characters (approximately)	X	43	5	Weighted Euclidean distance + array disorder	2.2 FAR, 4.6 FRR
Giuffrida et al. [11]	2014	8-9 characters	O	20	40 (approximately)	kNN () Manhattan weighted,; kNN () Manhattan scaled weighted	8
Antal and Szabó [12]	2015	10 characters	X	42	2/3 of data	Manhattan distance	12.9
Chang et al. [13]	2016	6-digit PIN; 8-digit PIN; 10-digit PIN	X	100	100	Statistical	23; 21; 16
Wu and Chen [14]	2015	8-digit PIN	O	100	500	SVM	0.556
Buschek et al. [15]	2015	6–8 characters	X	28	-	Probabilistic modeling	21.02
Dhage et al. [16]	2015	10 characters	X	15	10	Statistical	0.806
Bond and Awad [17]	2015	34 characters	X	25	-	Neural network	9.3
Teh et al. [18]	2016	4-digit PIN; 16-digit PIN	X	50/150	7	Gaussian estimation,; -score matching function,; standard deviation drift	7.57; 5.49