Preventing shoulder-surfing when selecting pass-images in challenge set

In this paper we present an experiment to detect the vulnerability of recognition based graphical authentication system to shoulder surfing. The experiment is conducted with teams of two, a user logging in to such as system and an observer who is trying to spot the user's pass-image. Our results suggest that inputting our choice via the keyboard is more secure than using a mouse. We use three different ways of clicking keyboard input and find that they are all approximate equally secure.