Inkblot Authentication

  • Adam Stubblefield
  • Dan Simon

MSR-TR-2004-85 |

In this paper we present a system that helps users select, remember, and differentiate strong passwords. Similar in nature to a Rorchach inkblot test, the system asks the user to form semantic associations with a set of randomly generated inkblot-like images. These image associations can then be used to authenticate the user. This approach is based on evidence from the psychological literature that suggests users will both choose dissimilar associations and retain their associations for long periods of time. We report on two user studies which show that inkblot based passwords are both memorable and high in entropy.