Evaluating the use of models of visual attention to predict graphical passwords
Liam M. Mayron, Mohammad N. Alshehri · 2014
In this work, we evaluate the use computational models of visual attention to predict graphical passwords. We compare the performance of three models: the Itti-Koch-Niebur model, Graph-Based Visual Saliency (GBVS), and the Image Signature. The visual attention maps generated by the models are compared to ground truth user password location selections. Although the Itti-Koch-Niebur model and GBVS have previously been applied individually to the challenge of predicting graphical passwords, this is, to our knowledge, both the first work to evaluate the image signature, and to compare the performance of different models of visual attention. This work demonstrates that, although all three models show potential for predicting graphical passwords, GBVS exhibits the overall best performance.