Abstract or Summary:
Graphical authentication holds some potential as an alternative to the ubiquitous password. Graphical authentication mechanisms typically present users with one or more challenge sets composed of a number of images: one target image surrounded by distractor images. Unfortunately, this means it tends to be more time-consuming than password entry and to alleviate this, we need to streamline the process as much as possible to maximize efficiency. The distractors must be chosen with care so as to ensure that users do not become confused by similarities with the target image. It is especially challenging to achieve this filtering with minimalist image types, such as hand-drawn doodles.

This paper explores the issues related to filtering the distractor images used in graphical authentication mechanisms using minimalist images. We present an algorithm for automatically classifying minimalist images in terms of visual similarity. The principles outlined here can also be used to assess the similarity of other minimalist image types such as signatures and handwritten numerals.