A Survey of Keystroke Dynamics Biometrics

Table 4

Comparison between research works with static authentication mode (short inputs).

Study	Data size	Latency	Input repetition	Input freedom	Method	FAR (%)	FRR (%)	EER (%)	Input sessions	Device freedom
[113]	33	FT	8	Yes	Mean, standard deviation	0.25	16.36	—	—	No
[61]	26	FT	30	Yes	Bayesian and minimum distance classifier	2.8	8.1	—	Once	No
[155]	24	FT	2	Yes	Perceptron algorithm	8	9	—	Yes	No
[9]	15	DT, FT	—	Yes	ART-2, RBFN, and LVQ	—	—	0	Yes	No
[103]	10	DT, FT	20	Yes	Inductive learning classifier	9	10	—	—	No
[84]	21	DT, FT	150–400	Yes	Autoassociative multilayer perceptron, SVM	0	0.814	—	—	—
[134]	100	DT	100	No	Genetic algorithm	—	—	95*	Yes	—
[21]	41	DT, FT	5	No	Random forest decision tree	—	—	2	Yes	Yes
[42]	30	DT, FT	10	No	Sequence alignment algorithms	0.15	0.2	0.35	Yes	—
[17]	21	DT, FT	—	Yes	K-means, euclidian	—	—	3.8	Yes	—
[86]	100	DT, FT	—	No	Multilayer perceptron	1	8	—	Once	—
[39]	41	DT, FT	30	Yes	Gaussian mixture modeling	4.3	4.8	4.4	Yes	Yes
[89]	100	DT, FT	6	No	Bayesian, Euclidean, hamming distance	—	—	6.96	Yes	No
[67]	1254	DT, FT	20	No	Mean, standard deviation	16	1	—	Once	Yes
[96]	16	DT, FT	5	No	Bayesian, Euclidean	—	—	4.28	Yes	—
[108]	25	DT, FT	30	Yes	Gauss, Parzen, K-NN, K-mean	—	—	1	—	—
[123]	51	DT, FT	50	No	Manhattan distance	—	—	7.1	Yes	No
[93]	100	DT, FT	12	No	Support vector machine	—	—	15.28	Yes	No
[3]	100	DT, FT	10	No	Gaussian PDF, direction similarity measure	—	—	1.401	Once	Yes
[138]	117	DT, FT	5	Yes	Support vector machine	—	—	11.83	Once	No

* Indicates performance measurement in terms of accuracy, similar but inverse to EER where value closer to 100% indicates better performance.