ABSTRACT
How many passwords and PINs do you have to remember just to manage your affairs
on the Internet? In fact, many deficiencies of password authentication systems arise in
consequence of regular working conditions of human memory. If one needed not to remember
passwords, they could be very secure indeed.
Traditionally, Information Security approaches to the password problems seem to have
focused on the technological aspects. Besides, the user has been considered the weakest link
in the security chain, indicating that the human factors end up jeopardizing the security that
the technology is supposed to enhance. Hence, it is necessary to better understand the human
factors involved in authentication, so that these factors can be accommodated into both new
and existing systems.
Therefore, this dissertation reports two empirical studies. The first study describes a
survey conducted in order to identify the main factors that hinder password recall. For this
study, 263 male and female participants were interviewed. Participants’ ages ranged between
18 and 93, and education level ranged from grade school to graduate degree. The results
indicated that, regardless of age or educational level, the number of password uses (in average
5.38 passwords per user) is the factor that influences memory performance the most. Thus,
better-educated users, for owning more passwords, were more prone to password forgetting
and mix-ups. Contrary to the expectations, the effect of cognitive decline (due to the aging
process) on password memory was not observed. In sum, the necessity of memorizing strong
passwords, by ignoring the natural working conditions of human memory, generates revenge
effects: habits that jeopardize the very reason for using passwords.
In the second study, we intended to explore ideas based on Cognitive Psychology
aiming at enhancing memory performance in password usage. By means of two experiments,
the study investigated the effect of elaborative rehearsal and of cue support as an aid to
password recall, with the goal of encouraging the generation of strong and memorable
passwords. Experiment 1 evaluated the effect of elaborative rehearsal and of cue support on
password composition, length, and security potential, as compared to a control group.
Password recall was tested in two occasions, 5 minutes after password generation and after a
week interval. The results from Experiment 1 indicate that people tend to observe only the
requirements that are somehow enforced, in spite of the instructions. With Experiment 2, we
sought to evaluate the memory performance after a longer delay, that is, five weeks after
password generation. Group effects (experimental vs. control) on password recall were not
observed. However, a possible confound was identified, the spacing effect, caused by the 5
minute login which, according to memory studies, favors the encoding process. In both
experiments, recall levels were higher than expected. In addition, the errors from the
unsuccessful login attempts were carefully analyzed and categorized. The observed error
types suggest that, oftentimes, the users do remember the gist of the password, but forget the
details of the format in which the password was coded.
This work concludes with considerations about the main findings described in the two
studies, as well as their potential implications. Moreover, in an attempt to bridge the gap
between the world of technology and the world of its human users, whose interaction has
often been overlooked, we point out suggestions for future investigations and limitations of
the reported studies.
Keywords: passwords, memory, information security, cued recall, elaborative rehearsal
NÚMERO DA ÁREA DO CNPq:
7.07.00.00-1 : Psicologia
7.07.02.00-4 : Psicologia Experimental
7.07.06.00-0 : Psicologia Cognitiva