LuTBA: Lightweight and Updatable Touch-Based Authentication Scheme for Mobile Devices

Summary

LuTBA is a touch-based authentication scheme for mobile devices that identifies users by measuring similarity between inputs and templates rather than training user-specific models during enrollment. Evaluated on two real-world datasets, it reaches optimal performance with 10 enrolled samples, one-eightieth of the training samples used by baselines, and reduces the mean equal error rate by 2.08%-24.83%. Its dynamic update mechanism handles behavior drift, and it attains an average equal error rate of 8.2% against imitation attacks at a window size of 10.