Exploring the Guessability of Image Passwords

This paper presents a user study with 70 participants using four different image types: Mikon; doodle; art; object, to examine guessability of image passwords to textual descriptions The study reported in this paper will examine, whether a written (textual) copy made by an account holder for recalling the target images forming a password in subsequent use, could be used for successful authentication by an attacker. The statistical test results of the study revealed that object images had the highest guessability, followed by the doodle images. The guessability of Mikon images was either equal to or slightly higher than the art images. The results also showed that the subjects described the elements in the target image, while making a written copy of their passwords for subsequent use, in the given experimental set-up. Based on the results obtained, we recommend the use of art images as passwords in image authentication