Poster: Image Suitability for Graphical Passwords

This work proposes a metric for determining the suitability of guiding images for graphical passwords. This measurement is intended to favor images that encourage users to select memorable, but non-obvious password click points. The metric was developed by analyzing thousands of passwords on a small but varied image database. We found that saliency covered the most frequently clicked portions of an image. Our results show that our model accurately predicts which images will score highest on both memorability and strength, thus providing an indicator as to which images are more suitable as guiding images for graphical passwords.