A Robust Continuous Authentication System Using Smartphone Sensors and Wasserstein Generative Adversarial Networks

Table 1

Mean EER (%) of real data and augmented data for the four classifiers at different data sizes.

Classifiers	Real data				Real data + Augmented data
	25%	50%	75%	100%	25%	50%	75%	100%
KNN	21.28	15.54	13.65	12.78	12.18	11.69	9.08	8.57
OCSVM	22.18	13.58	9.87	10.56	25.13	11.83	3.68	4.02
RF	19.25	10.58	9.67	8.86	7.89	6.94	5.48	4.84
DT	15.89	12.89	9.68	7.73	9.86	7.52	6.83	5.94

The mean EER is reduced by 48% after the WGAN augmentation method is used, and the mean EER is 3.68%, it performs the best performance.