Visual Pattern Recognition: Identity Authentication Technology Based on Eye Movement Patterns

Abstract

Existing identity authentication techniques are primarily based on passwords or physical characteristics and are therefore prone to theft and forgery. This can lead to potential security risks for identity recognition. In this paper, we propose a new identity authentication technique based on reading eye movements by introducing eye tracking technology. In this approach, a deep neural network based on multi-input architecture is used to construct a computational model. The model uses the text sequence, the fixation point sequence, and the linguistic feature as inputs, and output the recognition by a neural network. The simulation results showed a recognition accuracy of 89.9%. The eye tracking technique developed in this study may be used as a new and efficient identity authentication method. The data set used in the experiment and the associated codes have been released on GitHub.