Automatic registration of user drawn graphical passwords

This paper reports on a recognition based graphical authentication system where the users provide their pass images as simple drawings. Previous such systems have relied on a human administrator to register the images with the system, which limits their scale. We have replaced the administrator with software that both guides the user through the registration process and then automatically corrects any errors in the image files submitted to the system. We also investigate two ways of implementing these systems, by drawing on paper and scanning in the images, and using a computer paint program to create the images. We conducted a user study with 40 effective users. Our software was able to correct all the errors made by the users, extracting usable images in all cases. We also found that our users strongly preferred using the paint system to the system that required the use of a scanner. This research has implications for authentication with mobile phones, where the camera can replace the scanner and finger drawing apps replace the paint program..