With the popularity of mobile business, it becomes essential for mobile devices to check regularly whether the current user is the same one who unlock the device. The challenge lies on how to authenticate users continuously without influence on device usability. To guarantee both security and usability of mobile devices, we propose TDSD, a touch dynamic and sensor data based approach. It depicts user behavioral features in more comprehensive aspects and identify individuals with higher certainty. We utilize six classifiers and combine them together to form a stronger model by a compound-voting mechanism. Our experiments on real world dataset show a satisfactory performance of TDSD. Based on a small sample from 30 participants, compound-voting mechanism improve the accuracy of TDSD from about 85.00% to 100.00% and drop its false accept/reject rates to 0.00%.
Lin, Hong; Liu, Jiafen; and Li, Qing, "TDSD: A Touch Dynamic and Sensor Data Based Approach for Continuous User Authentication" (2018). PACIS 2018 Proceedings. 294.