Conference PaperPDF Available

Understanding Shoulder Surfing in the Wild: Stories from Users and Observers


Abstract and Figures

Research has brought forth a variety of authentication systems to mitigate observation attacks. However, there is little work about shoulder surfing situations in the real world. We present the results of a user survey (N=174) in which we investigate actual stories about shoulder surfing on mobile devices from both users and observers. Our analysis indicates that shoulder surfing mainly occurs in an opportunistic, non-malicious way. It usually does not have serious consequences, but evokes negative feelings for both parties, resulting in a variety of coping strategies. Observed data was personal in most cases and ranged from information about interests and hobbies to login data and intimate details about third persons and relationships. Thus, our work contributes evidence for shoulder surfing in the real world and informs implications for the design of privacy protection mechanisms.
Content may be subject to copyright.
Understanding Shoulder Surfing in the Wild:
Stories from Users and Observers
Malin Eiband, Mohamed Khamis, Emanuel von Zezschwitz, Heinrich Hussmann, Florian Alt
Media Informatics Group, LMU Munich, Germany
{malin.eiband, mohamed.khamis, emanuel.von.zezschwitz, hussmann, florian.alt}
Research has brought forth a variety of authentication sys-
tems to mitigate observation attacks. However, there is little
work about shoulder surfing situations in the real world. We
present the results of a user survey (N=174) in which we in-
vestigate actual stories about shoulder surfing on mobile de-
vices from both users and observers. Our analysis indicates
that shoulder surfing mainly occurs in an opportunistic, non-
malicious way. It usually does not have serious consequences,
but evokes negative feelings for both parties, resulting in a va-
riety of coping strategies. Observed data was personal in most
cases and ranged from information about interests and hob-
bies to login data and intimate details about third persons and
relationships. Thus, our work contributes evidence for shoul-
der surfing in the real world and informs implications for the
design of privacy protection mechanisms.
Author Keywords
Shoulder Surfing; Privacy; Mobile Devices
ACM Classification Keywords
H.5.2 Information Interfaces and Presentation: User
Interfaces—Input devices and strategies; K.6.5 Computing
Milieux: Security and Protection—Authentication
At the time of submission of this paper, at least 4640 aca-
demic publications indexed on Google scholar are concerned
with research on shoulder surfing1. The vast majority of these
articles (about 4000) have been published since 2007, the year
the iPhone entered the market. Not only since then, shoul-
der surfing – that is the act of observing other people’s infor-
mation without their consent (see Figure 1) – has served as
a fundamental motivation behind much of the work that has
been conducted in the area of usable privacy and security. In
particular, there is a plethora of work on authentication sys-
tems that aim to mitigate shoulder surfing on mobile devices
(e.g., [3, 8, 11, 18, 39, 45]).
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for profit or commercial advantage and that copies bear this notice and the full citation
on the first page. Copyrights for components of this work owned by others than the
author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior specific permission
and/or a fee. Request permissions from
CHI 2017, May 06 - 11, 2017, Denver, CO, USA
Copyright is held by the owner/author(s). Publication rights licensed to ACM.
ACM 978-1-4503-4655-9/17/05$15.00
Figure 1. A shoulder surfing situation in a cafe.
At the same time, surprisingly little is known about the phe-
nomenon of shoulder surfing itself. To date, there are no de-
tailed investigations of shoulder surfing incidents and their
real-world implications. In fact, findings from Harbach et
al. [20] in 2014 showed that only in eleven out of 3410 si-
tuations (0.3 %), smartphone users perceived a shoulder sur-
fing risk. Hence, one might wonder, how much of a threat is
shoulder surfing for the user?
To answer this question, this work contributes the results of an
exploratory survey (N=174) that collects actual stories from
both perspectives: users (the person being shoulder surfed)
and observers (the shoulder surfer). Stories are not restricted
to authentication, but focus on visual privacy in general.
The analysis of the stories revealed that shoulder surfing was
mostly casual and opportunistic. There were no stories indi-
cating observation out of malicious intent and/or with tech-
nical equipment. Shoulder surfing was most common among
strangers, in public transport, during commuting times, and
involved a smartphone in almost all cases. Observations un-
covered a broad range of mostly personal content, such as in-
formation about a user’s interests, hobbies, relationships, sex-
ual preferences, third persons, and login data. Except for two
cases, users did not report serious consequences of shoulder
surfing. However, both users and observers expressed nega-
tive feelings in the respective situation, such as embarrass-
ment and anger or guilt and unease. Users reacted with vari-
surfing%22,accessed 2016/09/20.
ous coping strategies, like turning the device away, shielding
content with their body, or adapting their way of interaction.
However, shoulder surfing often went unnoticed: Users were
aware of it in only 7 % of the incidents reported by observers.
The contribution of this work is threefold: First, we provide
evidence for shoulder surfing in the wild by collecting ex-
periences from users and observers. Second, we identify and
discuss seven main findings that help the community to better
understand the nature and implications of shoulder surfing.
Third, we derive implications of these results on the design
of future privacy protection mechanisms.
The observability of knowledge-based authentication mecha-
nisms has been discussed for over a decade [35, 49]. As estab-
lished authentication methods like Android unlock patterns
[46] and text-based input [37] are very easy to observe, most
prior work focused on investigating and proposing alternative
authentication concepts which conceal the entered secret and
make the authentication process observation-resistant. A lite-
rature review indicates that improved observation-resistance
is usually achieved by visual overload (e.g., [18, 45, 40]), in-
direct input (e.g., [12, 49]), multiplexed input (e.g., [35, 44])
or by establishing a second, non-observable, communication
channel (e.g., [4, 11, 26, 36, 51]). That is, the presentation
style or interaction concept are usually modified in a way that
makes the input harder to observe. The performance of these
concepts is mostly evaluated in the lab where authentication
is performed by participants and observation attacks are sim-
ulated by the experimenter [50]. Hence, the real-world vulne-
rability of authentication methods and the real-world risks of
shoulder surfing have not yet been evaluated systematically.
However, shoulder surfing is not limited to the observation
of password input. The Ponemon Institute [23] investigated
the feasibility of shoulder surfing attacks in business office
environments and found that “sensitive information can be
displayed on laptops, tablets and smartphones, as well as in
paper documents that are left in plain sight on desks, prin-
ters and conference tables and at other office locations or out-
side meeting sites”. The authors report that 91 % of all at-
tacks were successful. While 12 % of the observed informa-
tion was indeed based on login credentials (e.g., passwords),
personal information about consumers and employees (28 %),
contact lists (17 %) and financial information (8 %) was also
frequently observed. Overall, 28 % of the attacks were per-
formed on unprotected computer screens. In another study
[21], 85 % of the participants acknowledged that they had
already observed sensitive information on computer screens
they were not authorised to see. In addition to direct obser-
vation, threats to visual privacy have recently been discussed
in the context of surveillance cameras [38], corneal reflec-
tions [53], drones [48] and life logging cameras [22].
In contrast to the manifold work which was published to pro-
tect the authentication process, general visual privacy con-
cepts have not been in focus of research and related work
is quite sparse. Besides the use of physical privacy foils
which limit the angle of view [34], adaptive sensor-based ap-
proaches (e.g., [2, 7, 28, 54]) and static software-based ap-
proaches (e.g., [14, 41, 47]) have been discussed. Sensor-
based approaches are usually checking the user’s surrounding
and indicate bystanders. If potential observers are detected,
the user is either informed (e.g., [2, 7]) or the user interface
is adapted (e.g., [7, 28, 54]). Such adaptions imply reducing
the screen’s lightness or selectively hiding information [41].
In addition, individual user characteristics can be exploited to
improve visual privacy for specific content. Eiband et al. [14]
proposed to utilise the user’s handwriting to protect private
text-messages on mobile devices. Von Zezschwitz et al. [47]
suggested to use graphical distortion filters to protect personal
images stored in photo galleries of smart phones. Distortion
filters were applied so as to make the images’ content easy to
recognise for the user but hard to understand for observers.
Despite the effort of improving visual privacy and
observation-resistance of authentication mechanisms, the
real-world threat of shoulder surfing in everyday life is still
largely unknown. Harbach et al. [20] performed an experi-
ence sampling field study to evaluate the risk perception of
smartphone users in the wild. They reported that shoulder
surfing was rarely perceived as a relevant risk. However, the
authors acknowledged that critical environments (e.g., pub-
lic transport) might have been under-sampled. Trewin et al.
[43], who investigated the risk perception of users accessing
different kinds of specific data (e.g., company data, banking
information) on mobile devices, found that shoulder surfing
risks were frequently perceived. In line with this finding, Lit-
tle and Briggs [29] revealed that most users show stress re-
actions when viewing sensitive data in public environments.
Indeed, mobile devices are often used in public context and
provide a wide range of applications [5, 6] which may be ex-
posed to the danger of unauthorised observation.
We argue that it is crucial to understand the nature of everyday
shoulder surfing scenarios to inform the design of solutions
and to understand if shoulder surfing is restricted to certain
data types (e.g., authentication) or if it is a general problem.
We collected a large number of diverse experiences from both
users and observers to investigate shoulder surfing in the wild.
For this exploratory approach, we used an online survey2.
Design and Method
Shoulder surfing is a potentially sensitive topic, since con-
tent is observed without the other person’s knowledge or con-
sent. Prior work has shown that asking sensitive questions
in self-report should be done in an indirect and anonymity-
preserving way in order to minimise social desirability
bias [30, 31]. We also wanted to avoid a particular context
or use case to minimise recall bias as much as possible and to
capture diverse experiences.
We therefore decided to collect stories about shoulder surfing
based on the critical incident technique [16], which allows
for “generating a comprehensive and detailed description of
a content domain” [52]. The questionnaire was designed in
an iterative process and tested repeatedly in prestudies with
small participant samples. The critical incident was presented
as a sketch of an intentionally generic shoulder surfing situa-
tion among two stick-figures that we called “Cas” (the user)
and “Vic” (the observer), shown in Figure 2, alongside a short
description. We did not mention the term “shoulder surfing”
itself throughout the survey. We also avoided words and ex-
pressions with a negative connotation like “peeking” or “vic-
tim” to not give participants the impression that their be-
haviour is judged and thus possibly discourage them from
telling the truth. Instead, we used neutral terms like “loo-
king” and “user”. The stick-figures should allow to tell a story
anonymously, if a participant wanted to. We chose unisex
names instead of commonly used placeholders like “Alice”
and “Bob” because the results of the pretests suggested that
gender might influence recall. The pretests also indicated that
examples could sometimes not be completely avoided in or-
der to illustrate questions or statements. In these cases, we
tried to minimise bias by giving a very broad range of exam-
ples and included validity checks to identify participants who
copied text from the questions.
Questionnaire Structure
The questionnaire consisted of 1) a free text entry for parti-
cipants’ stories, 2) a section asking for specific details about
the story told in the first part, and 3) demographic questions.
In the first part, the shoulder surfing sketch and the descrip-
tion were displayed. Participants had to state whether they
knew of a situation like the one depicted. If their answer was
“Yes”, they were asked to report any real life experiences with
such an incident as detailed as possible via free text entry.
In the second part, we enquired about specific details, such as
the context of the situation, the type of content displayed on
the screen, the gender of both user and observer, etc. We also
asked participants to state the feelings and reactions of user
and observer and to optionally reveal their role (e.g., user or
observer). This allowed us to indirectly derive further insights
about the situation. We used multiple choice where appropri-
ate, but most of the questions allowed for free text entry.
In the last part, participants filled in demographic data and
had to indicate their honesty on a 5-point Likert scale. This
question was included as previous work suggested that self-
report of honesty helps identifying invalid data [46].
The survey was distributed in Egypt, Germany and the US.
It was framed as “Privacy on mobile devices” without men-
tioning the term “shoulder surfing” so as to minimise self-
selection bias in the sample. Participants were mainly re-
cruited via mailing lists and social media. More than two
thirds of the participants (70 %) were female. 75% were Ger-
man, 16 % Egyptian, the rest came from a variety of coun-
tries including the US, Bulgaria, India, Italy, Romania, Rus-
sia and South Korea. Participants’ age ranged from 16–57
years, with a mean of 25 years (SD=6 years). About two
thirds of the sample (67 %) were students. All data was stored
anonymously. Participants were compensated with vouchers
or credit points for their studies.
* Erforderlich
In this sketch you see Cas and Vic. Cas is using a mobile device
(like a smartphone or tablet) and is **not aware** of Vic looking
and seeing what’s on the screen of the device (e.g. text, pictures,
passwords/PINs, maps, videos, apps, games, websites etc.).
Do you know of a **real** situation in which this happened? Cas
and Vic could both be you or anyone else.
Privacy on mobile devices
1 von 1 03.01.2017 18:13
Figure 2. Screenshot of the first part of the online questionnaire.
Self-report is a common tool to gather a general under-
standing of user perceptions in diverse research areas and
established practice also in many security and privacy stu-
dies [1, 25]. However, we are aware that self-reported data is
susceptible to inaccurate statements, influence by wording or
recall bias [33]. For example, we cannot be sure whether an
incident in our survey was reported because it happened re-
cently or because participants felt it was serious. Moreover,
when asking people to self-report potentially sensitive ex-
periences, social desirability can be a source of error [42].
Also, since the study was posted online, our participant sam-
ple might have been biased by self-selection.
Although we cannot completely rule out any of these limi-
tations, we countered them in various ways: We followed an
indirect questioning approach allowing for anonymity. The
questions were designed in a careful, iterative way using mul-
tiple pretests. To identify invalid responses, we included va-
lidity checks such as a question asking for the honesty of the
participants, and manually inspected all collected data before
analysis as explained in detail in the next section. The framing
of the study also invited participants without shoulder surfing
experiences. The subset of those with shoulder surfing expe-
riences was identified post-hoc. Finally, research about shoul-
der surfing in the wild is not concluded with a single study and
that our results may not generalise, given our rather homo-
geneous sample. Still, we are confident that our exploration
gives directions for future work on specific issues.
A total of 176 participants completed our online survey. Each
statement was manually inspected before the coding pro-
cess. P102 and P157 both indicated 1=strongly disagree when
Code Count
Role: 174
– I was the observer 84
– I was the user 58
– I was neither 22
– I prefer not so say 4
– I don’t know of any incident 6
Relationship: 170
– strangers 126
– friends 11
– acquaintances 10
– colleagues 8
– family 3
– couple 3
– other 9
Gender: 151
– male observed female 44
– male observed male 38
– female observed female 37
– female observed male 32
Table 1. Role of participants, the relationship between user and ob-
server, and gender. Most stories were reported by observers, users and
observers were mostly strangers, and there was no correlation between
gender and role.
asked whether they were completely honest when answering
the questions. Their answers were therefore removed from the
data. We also checked whether participants copy-pasted from
the questions and whether their answers to the second part
of the questionnaire were coherent with the story told in the
first part. From the remaining 174 response sets, we derived
a codebook with 15 main categories from the collected an-
swers to the open-ended questions. Each category was further
refined by up to 21 subcategories, yielding 160 categories in
total. A randomly chosen subset of 10 % of the answers was
then coded independently by two raters. Agreement among
raters was calculated using Cohen’s κ[10] and Byrt et al.’s
prevalence-adjusted bias-adjusted κ(PABAK) [9], which are
appropriate indices given two raters and nominal data [13].
Statements could belong to more than one category. For in-
stance, the statement
“In public places and on public transport many people
are not aware that what they’re doing on their phone is
quite visible to people [...] nearby [...]” (P174)
was coded as Location: public open spaces and Location:
public transport. For this reason, comparisons were reduced
into 2×2 contingency tables based on whether a code was
present or absent in the report of each rater. Both κand
PABAK were then calculated for each of the 160 cate-
gories. For κ, interrater agreement was “substantial” to “(al-
most) perfect” [27] for 70 % of the subcategories. κranged
from 0.00 (e.g., Location: narrow, crowded places) to 1.00
(e.g., Time of day: evening). The range of values was smaller
for PABAK, where agreement ranged from 0.62 (Motivation
Observer: I don’t know) to 1.00 (e.g., Device: smartphone).
For PABAK, “substantial” to “(almost) perfect” agreement
was reached for 98 % of the categories. We assume that the
Code Count
Location: 193
– public transport 130
– theatre hall / lecture hall 13
– at work / university 10
– cafe, restaurant, bar 9
– narrow / crowded places 8
– public open spaces 7
– other 16
Activity: 189
– on the way 72
– commuting 69
– working / studying 16
– other 32
Time of day: 158
– morning 57
– evening 40
– afternoon 34
– midday 13
– all day 14
Device: 175
– smartphone 157
– tablet 8
– laptop 7
– ebook reader 3
Table 2. Context of shoulder surfing in everyday life. Shoulder surfing
mostly happened in public transport, during commuting time, and in-
volved a smartphone.
broad κrange was caused by effects of prevalence in the
respective categories, a known problem with κ[15]. There-
fore, no categories were excluded. All remaining discrepan-
cies were discussed until a consensus was reached. Assuming
that the final agreement generalises [19], half of the remai-
ning statements was then coded by the first, the other half by
the second rater.
The following sections present the insights gathered from the
analysis of the cleaned data, that is the remaining 174 re-
sponse sets, based on the coding. The percentages are cal-
culated based on the total number of collected code instances
in the respective category. Participants’ quotes are translated
to English where necessary.
Users and Observers
Table 1 shows the code instances related to the participants’
role and the relationship between user and observer. Six par-
ticipants (3.4 %) stated in the first part of the questionnaire
that they did not know of a shoulder surfing incident. 32 sto-
ries (18.4 %) were reported by participants who preferred to
remain anonymous or stated having been neither user nor ob-
server, but had, for example, witnessed the incident as a by-
stander. In 142 out of 174 cases, we could identify the par-
ticipants role (user or observer): 58 of the collected stories
(33.3 %) were told from the user perspective, 84 (48.3 %)
from the observer perspective. Only six out of these 84 ob-
servers (7.1 %) were noticed by the person they observed.
In most cases (74.1 %), user and observer were strangers, in
about a quarter (25.9 %) observer and user were acquainted.
These cases included shoulder surfing among friends (6.5 %),
colleagues (4.7 %), family members (P4, P163, P170; 1.7 %)
and couples (P19, P93, P94; 1.7 %), or incidents in which
the observer was the participants’ boss (P143, P153; 1.2 %).
There was no correlation between gender and role.
The code instances regarding contextual information are
shown in Table 2. Shoulder surfing mostly happened in public
transport: More than two thirds (67.4 %) of our participants
referred to a situation that occurred while travelling on the
subway, bus, train, etc., which is linked to the finding that user
and observer were mostly strangers. For example, P1 said that
she had “often tried to look at the phones of strangers in the
train”. P76 generalised this statement:
“Such a thing happens in public transport in particu-
lar. People are deliberately or unconsciously observed
by others while having, e.g., their smartphone in their
hand. This goes from quick glances to real staring.
P161 had a similar impression:
“In the subway [...] Everyone next to you or behind you
can look [at] your phone, you can look at [everyone’s]
phone there from the correct position.
Most commonly, user and observer were sitting side by side
(33.9 %). In 36.5 % of the cases user and observer had been
commuting. For instance, P8 said that she “[commutes] fre-
quently by train, where you often see people looking at the
phones of strangers and reading along.This is supported
by the fact that shoulder surfing happened mostly during the
times of the day when people typically commute: in the mor-
ning (36.1 %), afternoon (21.5 %) and evening (25.3 %).
Shoulder-surfed device were mainly smartphones (89.7 %).
Only few participants reported incidents involving a tablet
(4.6 %), laptop (4.0%) or ebook reader (1.7%). 21.8% of the
participants mentioned that shoulder surfing occurs regularly.
Table 3 shows the code instances related to the observed
data. The content observers saw on a display was mostly text
(46.6 %), followed by pictures (24.1 %) and games (12.6 %),
and was personal in the majority of the incidents (83.6 %).
Most commonly observed was communication via instant
messaging such as WhatsApp (41.8 %), followed by Face-
book (17.5 %), email (7.9 %) and news (7.4 %).
The insights observers gained from other people’s devices
were broad-ranging: “Resolution and font size as well as the
orientation of the device made it possible to follow everything
that was on the device and happening there [...]” (P50). Ob-
servers uncovered details about a user’s interests and hobbies,
weekend plans, work or studies, online shopping, appoint-
ments, and in some cases even about his or her sex life and
sexual preferences. In most cases, shoulder surfing revealed
information about a user’s relationships (34.3%), including
data about third persons. For example, P6 was able to see
someone’s apartment in a video chat.
Code Count
Content (general): 253
– text 118
– pictures 61
– games 32
– credentials 15
– videos 8
– other 19
Content (concrete): 189
– instant messaging 79
– Facebook 33
– email 15
– news 14
– PIN 9
– password 4
– online shopping 4
– unlock pattern 3
– other 28
Uncovered Information: 171
– personal: 143
– relationships / third persons 49
– interests / hobbies 23
– plans 17
– credentials 12
– work / studies 8
– sex life 5
– general 18
– other 11
– non-personal: 23
– games 15
– other 8
– other 5
Table 3. Type of content and information observed. Content was mostly
text in the form of instant messages, and personal. There was a broad
range of information uncovered by shoulder surfing.
In 5.9 % of the incidents, participants reported shoulder sur-
fing of authentication data (PINs, unlock patterns, passwords
and user names). For instance, take P173’s statement:
“One time I was on the train when I was opening my
phone using the pattern [...] which is provided as secu-
rity feature in [Android], when I noticed this guy trying
to figure out my pattern. When I asked him [did] you see
what I drew, he said ‘I think I know your pattern’ ”.
Moreover, the stories suggest that non-personal information
revealed details about the user. From the source the user read
news from, P11 and P23 uncovered information about politi-
cal views and interest in particular countries and topics. P14
reported learning “what kind of books” a neighbouring flight
passenger reads.
Motivation for Shoulder Surfing
The code instances regarding the observers’ motivation can
be found in Table 4. Our analysis indicates that observers
rarely shoulder surfed out of reasons other than curiosity and
boredom (33.6 % each). P82, for instance, said that “[it hap-
Code Count
Motivation Observer: 134
– curiosity 45
– boredom 45
– inadvertently 12
– device was in line of sight 8
– reaction to stimulus 6
– habit 4
– other 14
Table 4. Reasons for shoulder surfing. Observers acted out of curiosity
and boredom in most cases.
pened out of] boredom, because I cannot read a book while
standing”. Similarly, P19 stated that she was “[bored], [since
my] own smartphone was out of battery”. P1 admitted that
she shoulder surfed “to compare [my] own life with the life
of others because [I] don’t have many friends and [do] not
know what other people do”. Only few observers shoulder
surfed deliberately and regularly, for example:
“I always look at displays when you get the opportunity,
you get to know many things.(P10)
“[...] I like to read [other people’s WhatsApp chats]. Ro-
mantic drama, shopping lists, film recommendations ...
wherever you look!” (P43)
In other cases, shoulder surfing occurred inadvertently (e.g.,
P34, P40, P71) or as response to a stimulus such as light or
sound coming from the device (e.g., P20, P68).
There were no cases in which our analysis indicated shoulder
surfing with technical equipment or out of malicious intent,
such as getting a particular piece of information from the user
as leverage for the future.
Participants were asked to state both the user’s and the ob-
server’s feelings in the shoulder surfing situation. User fee-
lings were reported 41 times, observer feelings 90 times. The
code instances are shown in Table 5.
User Feelings
Our analysis indicates that feelings of users who noticed be-
ing observed were negative in the majority of cases (90.2 %),
among them anger, unease, pressure, and embarrassment. For
example, P116 described the following incident:
“I just had googled sex toys. Then I went to my pa-
rents’ bowling night. I then wanted to google something
quickly and my phone froze after the first letter and at
exactly that moment, the person sitting next to me at the
table looked at my display. There was written ‘inflatable
dildo’. One of the most embarrassing moments in my
life [...] I wished the ground would open and swallow
me up.
Other participants stated that users felt “violated in [their]
privacy” (P11), “harassed” (P15), “that someone is not
minding [their] own business” (P148), “very bad and uncom-
fortable” (P159), or “vulnerable” (P173).
Code Count
User’s Feelings: 41
– positive feelings: 1
– amused 1
– negative feelings: 37
– observed / spied on 13
– uneasy / embarrassed 12
– harassed / pressed 6
– angry / outraged 6
– other 3
Observer’s Feelings: 90
– positive feelings: 13
– amused 8
– superior 4
– proud 1
– negative feelings: 24
– uneasy 10
– guilty /ashamed 10
– annoyed 4
– curious / interested 26
– neutral 13
– bored 7
– other 7
Table 5. Feelings of users and observers. Shoulder surfing evoked nega-
tive feelings for both parties.
Moreover, participants expressed negative feelings through
criticism of the observer’s behaviour, for example,
P156: “[He felt] Hopefully like an idiot”, and P173: “[He
was] proud of himself that nosey bastard!” P166 summarised
her thoughts as follows:
“People [...] [looking] at your phone and [...] [reading]
your conversations or [focusing] on anything you are
doing [...] is totally unacceptable.
Observer Feelings
Interestingly, in more than a quarter of the cases (26.7%),
participants described the observer’s feelings as negative, too.
For instance, P176 stated that he “felt bad. This was [another
person’s] device.and P163 remarked:
“My sister is on [WhatsApp] and I [had] nothing to do
at that moment, so I [looked] at her screen. [...] when I
[realised] it [was] so unmoral, I turned my eyes away..
In other stories, the observer’s feelings were described as “a
bit more ashamed as if [...] reading along another person’s
newspaper” (P59), “a little bit guilty” (P161), “not com-
fortable, like [they] would not accept the privacy” (P67),
or “very wrong” (P146). P95 said:
“[...] If I catch myself staring at someone’s display, I –
particularly when it comes to sensitive content – quickly
look away”.
However, 14.4 % of the feelings were positive, like amuse-
ment or pride. Also, some observers were unconcerned about
shoulder surfing other people’s devices in public:
“Since I don’t look at my phone a lot in the train [...] I
now and then look at the phone of my neighbour. I don’t
see anything reprehensible about it. If something is top
secret, you don’t handle it in public.(P45)
We asked participants to state both the user’s and the ob-
server’s reactions during the incident, given that the user no-
ticed the observer. User reactions were reported 64 times, ob-
server reactions 41 times.
User Reactions
The code instances related to the user reactions are shown
in Table 6. Although the observer was sometimes just ig-
nored (7.8 %), shoulder surfing provoked a reaction in most
cases (92.2 %). Reactions occurred in a variety of ways, either
emotionally (20.3 %) or proactively (71.9 %), for example, by
protecting content from the observer’s view. Coping strate-
gies included turning the display or the whole body away
from the observer (43.5 %), switching the device or the screen
off, or putting the device away (13 % each). Interestingly,
users tended to behave differently depending on their rela-
tionship to the observer: If the observer was known to them,
users did not react with obvious rejection (e.g., by hiding the
display with their hands), but rather subtly by changing their
way of interacting with the device (e.g., quicker scrolling).
Moreover, some participants stated being alert when using
their mobile device in public, like P68:
“Everytime I take the subway I look at people and, if they
play a game, secretly watch what they are playing. For
this reason I’m always watchful when I’m doing some-
thing on my [smartphone], and look up frequently to see
whether someone is looking.
Others do not access sensitive data at all when in public:
“[...] I don’t do things that require higher security when
I’m on the way (online banking, entering credit card de-
tails, etc.)” (P62)
“[In public transport] you often notice that people look
at other (or my) smartphones. In these situations, I often
use other, ‘harmless’ apps or even stop using my smart-
P108 and P166 expressed a certain helplessness:
“Afterwards it occurred to me that eventually everyone
could have observed [...] what I’m typing and on which
page I’m on, but I also wouldn’t have known how to keep
my input secret.(P108)
“[...] you can’t really prevent them [looking] except by
not using your phone.(P166)
Emotional reactions included angry looks and verbal com-
plaints about the observer’s conduct. However, participants
also stated positive experiences where shoulder surfing led to
a conversation between user and observer (e.g., about a game
the user played), or triggered a humorous response (P93):
[Addressing the observer] “Shall I send [the message]
like that?” – [Observer] “Yes, that’s fine.
Code Count
Reaction User: 64
– ignored the observer 5
– proactive reaction: 46
– turned display / body away 20
– put device away 6
– turned device off 6
– changed way of interacting 6
– hid display with hands 3
– avoided using sensitive data 2
– other 3
– emotional reaction: 13
– angry look 4
– talked to observer (positive) 4
– talked to observer (negative) 3
– humorous / relaxed 2
Table 6. User reactions to shoulder surfing. If they noticed being ob-
served, most users tried to protect their content from the observer and
to prevent further shoulder surfing.
Observer Reactions
Observers mostly simply looked away from the device with-
out further reaction (61 %). Some observers did not take no-
tice of the user’s reaction and kept on looking (12.2%). Oth-
ers engaged in a conversation with the user (14.6%).
Shoulder surfing did not have unpleasant or severe conse-
quences except for two cases. Take P104’s statement of the
following incident, for example:
“I sat in the bus on the way to a friend who lives in a
neighbourhood said to be less safe. Apart from me there
were only two young men in the bus who sat [...] directly
behind me [...] I looked for the address [of my friend
on Google Maps] [...] They saw this and then started
talking about where I might be going. When getting off
the bus, they [...] said that they could show me the way
[...] it was getting dark already and [...] I had a bad fee-
ling [about it], since they made a somewhat aggressive
impression [...]”
In this section, we identify and discuss the main findings from
the analysis of the 174 stories collected in the online survey.
Shoulder Surfing is Real – But Goes Unnoticed
The great majority, 168 of our 174 participants (97%),
claimed to know of a shoulder surfing situation in everyday
life. Moreover, 21% of the participants mentioned that shoul-
der surfing occurs regularly, although we did not explicitly
ask them to report this detail. Leaving aside the incidents
reported by persons who were not directly involved in the
shoulder surfing incident, we collected 58 (33 %) stories from
users and 84 (48 %) stories from observers. Only six out of
these 84 observers stated being noticed by the user – about
7 %. Since stories by users could only be reported if the user
had noticed the observer, we argue that stories reported by
observers are better indicators of how often shoulder surfing
gets noticed. Hence, we conclude:
Finding 1: Shoulder surfing exists to a substantial amount
in real life. However, users are not aware of being ob-
served in the majority of cases.
Similar observations were made in related work through the
simulation of shoulder surfing attacks in offices [23].
Observers are not always Evil
The great majority of observers were strangers motivated by
curiosity and boredom (34 % each). Moreover, in many cases
(27 %) shoulder surfing went along with negative feelings
such as guilt or unease on the side of the observer. The in-
cidents we collected were all simple, one time observations
without technical equipment such as video cameras.
Finding 2: Observers are opportunistic and rarely act out
of reasons other than curiosity and boredom. Moreover,
they often associate their conduct with negative feelings.
While we cannot exclude the occurrence of premeditated at-
tacks in the wild, this indicates that real-world shoulder sur-
fing is mostly not based on malicious intent. Since the majo-
rity of observers in our survey were strangers (74 %), this
finding complements the work by Muslukhov et al. [32],
which suggests that insiders (e.g., friends and family) pose
a higher threat to users compared to strangers. In addition,
shoulder surfing may be a more serious risk in the business
context [23].
Personal Data Leakage Causes User Concern
The information uncovered by observers was personal in the
vast majority of cases (79 %) and affected a broad range of
private details, even of very intimate nature (e.g., sexual pre-
ferences). Although data leakage did mostly not have serious
consequences, almost all users who had noticed being shoul-
der surfed expressed negative feelings like anger or embar-
rassment (90 %) and tried to protect their data through a vari-
ety of coping strategies.
Finding 3: Shoulder surfing affects a broad range of per-
sonal information and evokes negative feelings on the side
of the user.
This suggests that shoulder surfing has a substantial influ-
ence on the user experience of mobile device usage in pub-
lic, which is in line with prior work by Little and Briggs [29]
who found that people respond with stress when their per-
sonal data is exposed publicly. It also indicates that mecha-
nisms are required which ensure visual privacy throughout
the interaction (e.g., [14, 47]).
Text, Pictures and Games are Most Observed Content
Content most commonly observed was text (47 %), followed
by pictures (24 %) and games (13 %). In particular, shoul-
der surfing mostly affected communication via instant messa-
ging, such as WhatsApp (42 %) and social network activities,
for instance, on Facebook (18 %).
Finding 4: Text is observed in most cases, followed by
pictures and games. This involves instant messaging and
social network activities in particular.
This correlates with mobile device usage investigated in pre-
vious work that showed that mobile devices are still commu-
nication tools in the first place [5]. Moreover, it again con-
firms the need for visual privacy-protection concepts.
Shoulder Surfing Affects Credentials
We collected 15 stories (9 %) in which shoulder surfers ob-
served login data. While in most of these cases (12 out of 15)
the user was entering an authentication PIN or unlock pattern,
there were also cases in which the user was authenticating on
a regular website with a user name and password.
Finding 5: Shoulder surfing puts authentication creden-
tials such as PINs, passwords and patterns at risk.
Previous work reported cases in which it was likely to be
shoulder surfed during authentication [20], and the feasibil-
ity of shoulder surfing in the workplace [23]. We report the
first evidence of real situations where authentication data was
shoulder surfed on mobile devices in the wild. In particu-
lar, this finding confirms assumptions made in previous work
about observation attacks (e.g., [11, 18, 45]).
Coping Strategies Depend on User/Observer Relation
As previously mentioned, most users did not notice the ob-
server. However, if they did, the majority of them took action
to prevent further shoulder surfing (46 out of 64 reactions –
72 %). We found that coping strategies tend to differ depen-
ding on the relationship between user and observer: If users
know the observer, they are likely to react in a more sub-
tle way (e.g. by scrolling quickly) than if the observer is a
stranger, where strategies are more obvious (e.g., turning the
phone away).
Finding 6: Users adapt their coping strategies against
shoulder surfing based on their relation to the observer.
This suggests that users are not only concerned about their
privacy, but also about the social implications of their reac-
tion, in particular if the observer is known to them. The im-
portance of social acceptability of security measures has been
highlighted in the literature [1]. Moreover, our finding is in
line with previous work, which reported that users are some-
times hesitant to unlock their phones in front of friends [32].
It is not Only the User’s Own Privacy at Stake
Personal information revealed by shoulder surfing was most
commonly related to a user’s relationships (49 out of 143
cases – 34 %). In more than a third of these cases (18 out of
49), the uncovered personal piece of information did not con-
cern the users themselves, but third persons they were com-
municating with (e.g., names, interests, insight into an apart-
ment via video chat in one case).
Finding 7: Shoulder surfing leaks personal information
about third persons through the content with which the
user interacts.
This is crucial in the business context and has been empha-
sised as challenge for the security of sensitive information
when people work while commuting [24].
From the findings discussed in the previous section, we de-
rive the following implications to guide researchers and prac-
titioners in designing feasible privacy protection systems and
improving the experience of using mobile devices in public:
1. Privacy protection should cover the broad range of
information that is leaked by shoulder surfing. While
Finding 5 supports the need for observation-resistant au-
thentication schemes, Finding 3, 4 and 7 show that due to
the popularity of instant messaging and social networks,
text and pictures are most prone to shoulder surfing in pub-
lic. Thus, research should extend its current focus on cre-
dentials to systems that protect other types of data. Exam-
ples for such systems are [14], [47] and [54].
2. Privacy protection should be lightweight, fast, and easy.
Since shoulder surfing is often casual without serious con-
sequences (Finding 2), users might not be willing to deal
with excessive overhead for the benefit of privacy. Prior
work has shown that users value usability more than se-
curity [17]. Systems for visual privacy thus have to be de-
signed and evaluated not only for the protection but also
for the usability they provide.
3. Privacy protection should not have to be initiated by the
user in a shoulder surfing situation. Finding 1 shows that
most users are unaware of being observed. In contrast to
approaches taken in prior work (e.g., [11, 45]), where pro-
tection is switched on and off by the user, this suggests that
automatically triggering protection or alerting the user (like
in [2]) is more effective. For this purpose, a system could
rely on contextual information to determine, for example,
whether a user is currently riding public transportation.
4. Privacy protection should be socially acceptable. Fin-
ding 6 suggests that, as with other security systems [1],
social acceptability has an impact on visual privacy pro-
tection. In order to avoid social implications, in particular
in cases in which the user is acquainted with the observer,
solutions need to be designed and evaluated with social ac-
ceptability in mind.
Shoulder surfing has motivated an abundance of studies and
systems in the past, in particular in the area of usable secu-
rity and privacy. To date however, the real-world relevance
and implications of shoulder surfing itself were still underex-
plored. This work contributes to the literature the results of
an online survey (N=174), in which we collected stories from
both users and observers about their experiences with shoul-
der surfing. The received reports provide insight into various
aspects of the nature of shoulder surfing. For example, ob-
servations are usually casual and opportunistic, but uncover a
broad range of personal information (e.g., authentication data,
interests, dating) and thus evoke negative feelings for both
users and observers. Users react with diverse coping strate-
gies that depend on their relationship to the observer. How-
ever, in most cases shoulder surfing goes unnoticed. Thus,
our findings improve understanding of shoulder surfing in the
wild and inform a set of implications for feasible privacy pro-
tection mechanisms. In future work, we want to 1) find ways
to quantify shoulder surfing in the field, 2) investigate the ex-
istence and the frequency of malicious shoulder surfing at-
tacks, and 3) examine usable ways to protect sensitive data
in public. Additionally, in-depth interviews could be a benefi-
cial next step to learn more about some of the specific issues
presented in this work.
1. Anne Adams and Martina Angela Sasse. 1999. Users
Are Not the Enemy. Commun. ACM 42, 12 (1999),
40–46. DOI:
2. Mohammed Eunus Ali, Anika Anwar, Ishrat Ahmed,
Tanzima Hashem, Lars Kulik, and Egemen Tanin. 2014.
Protecting Mobile Users from Visual Privacy Attacks. In
Proceedings of the 2014 ACM International Joint
Conference on Pervasive and Ubiquitous Computing:
Adjunct Publication (UbiComp ’14 Adjunct). ACM,
New York, NY, USA, 1–4. DOI:
3. Florian Alt, Stefan Schneegass, Alireza Sahami Shirazi,
Mariam Hassib, and Andreas Bulling. 2015. Graphical
Passwords in the Wild: Understanding How Users
Choose Pictures and Passwords in Image-based
Authentication Schemes. In Proceedings of the 17th
International Conference on Human-Computer
Interaction with Mobile Devices and Services
(MobileHCI ’15). ACM, New York, NY, USA, 316–322.
4. Andrea Bianchi, Ian Oakley, Vassilis Kostakos, and
Dong Soo Kwon. 2011. The Phone Lock: Audio and
Haptic Shoulder-surfing Resistant PIN Entry Methods
for Mobile Devices. In Proceedings of the Fifth
International Conference on Tangible, Embedded, and
Embodied Interaction (TEI ’11). ACM, New York, NY,
USA, 197–200. DOI:
5. Matthias B¨
ohmer, Brent Hecht, Johannes Sch¨
Antonio Kr¨
uger, and Gernot Bauer. 2011. Falling Asleep
with Angry Birds, Facebook and Kindle: A Large Scale
Study on Mobile Application Usage. In Proceedings of
the 13th International Conference on Human Computer
Interaction with Mobile Devices and Services
(MobileHCI ’11). ACM, New York, NY, USA, 47–56.
6. Barry Brown, Moira McGregor, and Donald McMillan.
2014. 100 Days of iPhone Use: Understanding the
Details of Mobile Device Use. In Proceedings of the
16th International Conference on Human-computer
Interaction with Mobile Devices & Services (MobileHCI
’14). ACM, New York, NY, USA, 223–232. DOI:
7. Frederik Brudy, David Ledo, Saul Greenberg, and
Andreas Butz. 2014. Is Anyone Looking? Mitigating
Shoulder Surfing on Public Displays Through
Awareness and Protection. In Proceedings of the
International Symposium on Pervasive Displays (PerDis
’14). ACM, New York, NY, USA, Article 1, 6 pages.
8. Andreas Bulling, Florian Alt, and Albrecht Schmidt.
2012. Increasing the Security of Gaze-based Cued-recall
Graphical Passwords Using Saliency Masks. In
Proceedings of the SIGCHI Conference on Human
Factors in Computing Systems (CHI ’12). ACM, New
York, NY, USA, 3011–3020. DOI:
9. Ted Byrt, Janet Bishop, and John B. Carlin. 1993. Bias,
Prevalence and Kappa. Journal of clinical epidemiology
46, 5 (1993), 423–429.
10. Jacob Cohen. 1960. A Coefficient of Agreement for
Nominal Scales. Educational and Psychological
Measurement 20, 1 (1960), 37–46.
11. Alexander De Luca, Marian Harbach, Emanuel von
Zezschwitz, Max-Emanuel Maurer, Bernhard Ewald
Slawik, Heinrich Hußmann, and Matthew Smith. 2014.
Now You See Me, Now You Don’t: Protecting
Smartphone Authentication from Shoulder Surfers. In
Proceedings of the SIGCHI Conference on Human
Factors in Computing Systems (CHI ’14). ACM, New
York, NY, USA, 2937–2946. DOI:
12. Alexander De Luca, Katja Hertzschuch, and Heinrich
Hußmann. 2010. ColorPIN: Securing PIN Entry
Through Indirect Input. In Proceedings of the SIGCHI
Conference on Human Factors in Computing Systems
(CHI ’10). ACM, New York, NY, USA, 1103–1106.
13. Michael E. Dewey. 1983. Coefficients of Agreement.
The British Journal of Psychiatry 143, 5 (1983),
14. Malin Eiband, Emanuel von Zezschwitz, Daniel
Buschek, and Heinrich Hußmann. 2016. My Scrawl
Hides It All: Protecting Text Messages Against Shoulder
Surfing With Handwritten Fonts. In Proceedings of the
2016 CHI Conference Extended Abstracts on Human
Factors in Computing Systems (CHI EA ’16). ACM,
New York, NY, USA, 2041–2048. DOI:
15. Alvan R. Feinstein and Domenic V. Cicchetti. 1990.
High Agreement but Low Kappa: The Problems of Two
Paradoxes. Journal of Clinical Epidemiology 43, 6
(1990), 543–549.
16. John C. Flanagan. 1954. The Critical Incident
Technique. Psychological Bulletin 51, 4 (1954),
17. Simson Garfinkel and Heather Richter Lipford. 2014.
Usable Security:History, Themes, and Challenges.
Morgan & Claypool. 124– pages. DOI:http:
18. Jan Gugenheimer, Alexander De Luca, Hayato Hess,
Stefan Karg, Dennis Wolf, and Enrico Rukzio. 2015.
ColorSnakes: Using Colored Decoys to Secure
Authentication in Sensitive Contexts. In Proceedings of
the 17th International Conference on Human-Computer
Interaction with Mobile Devices and Services
(MobileHCI ’15). ACM, New York, NY, USA, 274–283.
19. Kevin A. Hallgren. 2012. Computing Inter-rater
Reliability for Observational Data: An Overview and
Tutorial. Tutorials in Quantitative Methods for
Psychology 8, 1 (2012), 23–34.
20. Marian Harbach, Emanuel von Zezschwitz, Andreas
Fichtner, Alexander De Luca, and Matthew Smith. 2014.
It’s a Hard Lock Life: A Field Study of Smartphone
(Un)Locking Behavior and Risk Perception. In
Symposium On Usable Privacy and Security (SOUPS
2014). USENIX Association, Menlo Park, CA,
21. Brian Honan. 2012. Visual Data Security White Paper.
22. Roberto Hoyle, Robert Templeman, Denise Anthony,
David Crandall, and Apu Kapadia. 2015. Sensitive
Lifelogs: A Privacy Analysis of Photos from Wearable
Cameras. In Proceedings of the 33rd Annual ACM
Conference on Human Factors in Computing Systems
(CHI ’15). ACM, New York, NY, USA, 1645–1648.
23. Ponemon Institute. 2016. Global Visual Hacking
Experimental Study: Analysis. (2016).
global-visual- hacking-experiment- study-summary.
24. Iron Mountain. 2013. Protecting sensitive company
information from the commuter snoopers. (2013).
Company-News/News- Categories/Press-Releases/
25. Amy K. Karlson, A.J. Brush, and Stuart Schechter.
2009. Can I Borrow Your Phone?: Understanding
Concerns When Sharing Mobile Phones. In Proceedings
of the SIGCHI Conference on Human Factors in
Computing Systems (CHI ’09). ACM, New York, NY,
USA, 1647–1650. DOI:
26. Mohamed Khamis, Florian Alt, Mariam Hassib,
Emanuel von Zezschwitz, Regina Hasholzner, and
Andreas Bulling. 2016. GazeTouchPass: Multimodal
Authentication Using Gaze and Touch on Mobile
Devices. In Proceedings of the 2016 CHI Conference
Extended Abstracts on Human Factors in Computing
Systems (CHI EA ’16). ACM, New York, NY, USA,
2156–2164. DOI:
27. J. Richard Landis and Gary G. Koch. 1977. The
Measurement of Observer Agreement for Categorical
Data. Biometrics 33, 1 (1977), 159–174. DOI:
28. Shiguo Lian, Wei Hu, Xingguang Song, and Zhaoxiang
Liu. 2013. Smart Privacy-preserving Screen Based on
Multiple Sensor Fusion. IEEE Transactions on
Consumer Electronics 59, 1 (2013), 136–143. DOI:
29. Linda Little and Pam Briggs. 2009. Private
Whispers/Public Eyes: Is Receiving Highly Personal
Information in a Public Place Stressful? Interacting with
Computers 21, 4 (2009), 316 – 322. DOI:
30. Diogo Marques, Tiago Guerreiro, and Luis Carric¸o.
2014. Measuring Snooping Behavior with Surveys: It’s
How You Ask It. In CHI ’14 Extended Abstracts on
Human Factors in Computing Systems (CHI EA ’14).
ACM, New York, NY, USA, 2479–2484. DOI:
31. Diogo Marques, Ildar Muslukhov, Tiago Guerreiro, Lu´
Carric¸o, and Konstantin Beznosov. 2016. Snooping on
Mobile Phones: Prevalence and Trends. In Twelfth
Symposium on Usable Privacy and Security (SOUPS
2016). USENIX Association, Denver, CO, 159–174.
32. Ildar Muslukhov, Yazan Boshmaf, Cynthia Kuo,
Jonathan Lester, and Konstantin Beznosov. 2013. Know
Your Enemy: The Risk of Unauthorized Access in
Smartphones by Insiders. In Proceedings of the 15th
International Conference on Human-computer
Interaction with Mobile Devices and Services
(MobileHCI ’13). ACM, New York, NY, USA, 271–280.
33. Delroy L. Paulhus and Vazire Simine. 2009. The
Self-report Method. In Handbook of Research Methods
in Personality Psychology, Richard W. Robins, R. Chris
Fraley, and Robert F. Krueger (Eds.). Guilford Press,
34. George Probst. 2000. Analysis of the Effects of Privacy
Filter Use on Horizontal Deviations in Posture of VDT
Operators. Ph.D. Dissertation. Virginia Polytechnic
Institute and State University.
35. Volker Roth, Kai Richter, and Rene Freidinger. 2004. A
PIN-entry Method Resilient Against Shoulder Surfing.
In Proceedings of the 11th ACM Conference on
Computer and Communications Security (CCS’04).
ACM, New York, NY, USA, 236–245. DOI:
36. Hirokazu Sasamoto, Nicolas Christin, and Eiji Hayashi.
2008. Undercover: Authentication Usable in Front of
Prying Eyes. In Proceedings of the SIGCHI Conference
on Human Factors in Computing Systems (CHI ’08).
ACM, New York, NY, USA, 183–192. DOI:
37. Florian Schaub, Ruben Deyhle, and Michael Weber.
2012. Password Entry Usability and Shoulder Surfing
Susceptibility on Different Smartphone Platforms. In
Proceedings of the 11th International Conference on
Mobile and Ubiquitous Multimedia (MUM ’12). ACM,
Ulm, Germany. DOI:
38. Jeremy Schiff, Marci Meingast, Deirdre K. Mulligan,
Shankar Sastry, and Ken Goldberg. 2009. Respectful
Cameras: Detecting Visual Markers in Real-Time to
Address Privacy Concerns. Springer London, London,
65–89. DOI: 84882-301- 3_5
39. Stefan Schneegass, Frank Steimle, Andreas Bulling,
Florian Alt, and Albrecht Schmidt. 2014. SmudgeSafe:
Geometric Image Transformations for Smudge-resistant
User Authentication. In Proceedings of the 2014 ACM
International Joint Conference on Pervasive and
Ubiquitous Computing (UbiComp ’14). ACM, New
York, NY, USA, 775–786. DOI:
40. Desney S. Tan, Pedram Keyani, and Mary Czerwinski.
2005. Spy-resistant Keyboard: More Secure Password
Entry on Public Touch Screen Displays. In Proceedings
of the 17th Australia Conference on Computer-Human
Interaction (OZCHI ’05). Computer-Human Interaction
Special Interest Group (CHISIG) of Australia,
Narrabundah, Australia, 1–10. http:
41. Peter Tarasewich, Jun Gong, and Richard Conlan. 2006.
Protecting Private Data in Public. In CHI ’06 Extended
Abstracts on Human Factors in Computing Systems
(CHI EA ’06). ACM, New York, NY, USA, 1409–1414.
42. Roger Tourangeau and Ting Yan. 2007. Sensitive
Questions in Surveys. Psychological bulletin 133, 5
(2007), 859–883. DOI:
43. Shari Trewin, Cal Swart, Larry Koved, and Kapil Singh.
2016. Perceptions of Risk in Mobile Transaction. In
2016 IEEE Security and Privacy Workshops (SPW).
214–223. DOI:
44. Wouter van Eekelen, John van den Elst, and
Vassilis-Javed Khan. 2014. Dynamic Layering
Graphical Elements For Graphical Password Schemes.
In Proceedings of the Chi Sparks 2014 Conference: HCI
Research, Innovation, and Implementation. The Hague
University of Applied Sciences, The Hague, The
Netherlands, 65–73.
45. Emanuel von Zezschwitz, Alexander De Luca, Bruno
Brunkow, and Heinrich Hußmann. 2015a. SwiPIN: Fast
and Secure PIN-Entry on Smartphones. In Proceedings
of the 33rd Annual ACM Conference on Human Factors
in Computing Systems (CHI ’15). ACM, New York, NY,
USA, 1403–1406. DOI:
46. Emanuel von Zezschwitz, Alexander De Luca, Philipp
Janssen, and Heinrich Hußmann. 2015b. Easy to Draw,
but Hard to Trace?: On the Observability of Grid-based
(Un)Lock Patterns. In Proceedings of the 33rd Annual
ACM Conference on Human Factors in Computing
Systems (CHI ’15). ACM, New York, NY, USA,
2339–2342. DOI:
47. Emanuel von Zezschwitz, Sigrid Ebbinghaus, Heinrich
Hußmann, and Alexander De Luca. 2016. You Can’t
Watch This!: Privacy-Respectful Photo Browsing on
Smartphones. In Proceedings of the 2016 CHI
Conference on Human Factors in Computing Systems
(CHI ’16). ACM, New York, NY, USA, 4320–4324.
48. Yang Wang, Huichuan Xia, Yaxing Yao, and Yun
Huang. 2016. Flying Eyes and Hidden Controllers: A
Qualitative Study of Peoples Privacy Perceptions of
Civilian Drones in the US. Proceedings on Privacy
Enhancing Technologies 2016, 3 (2016), 172–190.
DOI: 0022
49. Susan Wiedenbeck, Jim Waters, Leonardo Sobrado, and
Jean-Camille Birget. 2006. Design and Evaluation of a
Shoulder-surfing Resistant Graphical Password Scheme.
In Proceedings of the Working Conference on Advanced
Visual Interfaces (AVI ’06). ACM, New York, NY, USA,
177–184. DOI:
50. Oliver Wiese and Volker Roth. 2015. Pitfalls of Shoulder
Surfing Studies. In NDSS Workshop on Usable Security
2015 (USEC’15). Internet Society, 1–6.
51. Christian Winkler, Jan Gugenheimer, Alexander
De Luca, Gabriel Haas, Philipp Speidel, David
Dobbelstein, and Enrico Rukzio. 2015. Glass Unlock:
Enhancing Security of Smartphone Unlocking Through
Leveraging a Private Near-eye Display. In Proceedings
of the 33rd Annual ACM Conference on Human Factors
in Computing Systems (CHI ’15). ACM, New York, NY,
USA, 1407–1410. DOI:
52. Lorette K. Woolsey. 1986. The Critical Incident
Technique: An Innovative Qualitative Method of
Research. Canadian Journal of Counselling 20, 4
(1986), 242–254.
53. Yi Xu, Jared Heinly, Andrew M. White, Fabian
Monrose, and Jan-Michael Frahm. 2013. Seeing Double:
Reconstructing Obscured Typed Input from Repeated
Compromising Reflections. In Proceedings of the 2013
ACM SIGSAC Conference on Computer &
Communications Security (CCS ’13). ACM, New York,
NY, USA, 1063–1074. DOI:
54. Huiyuan Zhou, Vinicius Ferreira, Thamara Alves,
Kirstie Hawkey, and Derek Reilly. 2015. Somebody Is
Peeking!: A Proximity and Privacy Aware Tablet
Interface. In Proceedings of the 33rd Annual ACM
Conference Extended Abstracts on Human Factors in
Computing Systems (CHI EA ’15). ACM, New York,
NY, USA, 1971–1976. DOI:
... This is because they are often required as fallback schemes for biometric authentication, or because in some cases users prefer to use them to avoid sharing biometric data with third parties [21]. Driven by the need to improve the security of PINs, a simple approach is to randomize the layout of the keypad to make PINs more secure against common side channel attacks, such as shoulder surfing [9,11] and thermal attacks [1-3, 5, 8]. ...
... Threat modeling is often defined as the formal process of identifying, documenting, and mitigating security threats to a system [20]. Human-centered threats involve social engineering attacks [15], shoulder surfing attacks [9,11], thermal attacks [1,2,5], smudge attacks [7,24], just to name a few. Designing authentication schemes against multiple threats is challenging, mainly because protecting users against all security and privacy threats is close to impossible [18]. ...
... Designing authentication schemes against multiple threats is challenging, mainly because protecting users against all security and privacy threats is close to impossible [18]. One of the most common threat models in the broader HCI field is shoulder surfing [11], where bystanders observe a user during their authentication. In contrast works that protect against observation attacks, Aviv et al. [7] and Abdelrahman et al. [1] respectively investigated how smudge attacks (i.e., oily residues) and thermal attacks (i.e., heat traces on a screen) impact the security of traditional user authentication on mobile devices. ...
... Studying users' awareness of potential threats, e.g., thermal attacks, is important as previous work showed that users only apply protection mechanisms against risk they are aware of [11]. Our study shows that many users are aware of smudge attacks [5], shoulder surfing attacks [8], and what thermal cameras are. However, most of our participants reported to not consider thermal attacks as a realistic threat in their daily life. ...
... The second stream of research that is related to our work studies the security and privacy risk awareness of users. Researchers used surveys to investigate risk awareness of shoulder surfing [8], internet usage [10], password composition [28], mobile device usage [7,27], and wearable computing [6,9]. Overall, everyday users are often unaware of security and privacy risks associated with technology usage. ...
... The high awareness of shoulder surfing and the high adoption of precautions designed to prevent it shows that shoulder surfing is a risk which users perceive as relevant and protect themselves against. This is understandable, as shoulder surfing has been found to be a privacy threat that is common in everyday life [8], and so our participants are likely to have experienced shoulder surfing themselves. However, we also found that some participants were aware of risks but showed a lack of concern for them by not adopting precautions to mitigate them. ...
... Shoulder surfing through direct observation does not require special knowledge, since it is only a gaze at a person's device. Furthermore, shoulder surfers could be anyone, such as strangers, family members, friends, colleagues, or even intimate partners [13,33,36]. The ease of executing this attack and the fact that anyone could be a shoulder surfer makes shoulder surfing an ubiquitous threat. ...
... The ease of executing this attack and the fact that anyone could be a shoulder surfer makes shoulder surfing an ubiquitous threat. Several investigations in the literature underpin the existence of shoulder surfing in people's daily lives [13,36,43]. ...
... They found that many users are concerned about unauthorized access by friends and other "insiders". More generically, and most relevant to our work, is a shoulder surfing investigation by Eiband et al. [13] which provided the first evidence of shoulder surfing incidents in the real world. The study collected 174 shoulder surfing stories through a one-time online survey. ...
... Eiband et al. [7] conducted a survey study, collecting user stories of shoulder surfing situations. The focus was on shoulder surfing in general. ...
... First, they can select the category of the current location. Categories are derived from Eiband et al. [7] and include home, public transport, theater or lecture hall, work or university, cafe or restaurant or bar, crowded place, public space, and others. Besides the image, we presented the address, extracted from the GPS coordinates, at the moment of unlock to support the assessment. ...
... We categorized authentication events based on location (Figure 4-left). We thereby use the same categories as Eiband et al. [7]. Most unlock events occurred at home (50%), followed by participants' workplace or, in the case of students, at University (about 25%). ...
Full-text available
This paper contributes to our understanding of user-centered attacks on smartphones. In particular, we investigate the likelihood of so-called shoulder surfing attacks during touch-based unlock events and provide insights into users’ views and perceptions. To do so, we ran a two-week in-the-wild study (N=12) in which we recorded images with a 180-degree field of view lens that was mounted on the smartphone’s front-facing camera. In addition, we collected contextual information and allowed participants to assess the situation. We found that only a small fraction of shoulder surfing incidents that occur during authentication are actually perceived as threatening. Furthermore, our findings suggest that our notions of (un)safe places need to be rethought. Our work is complemented by a discussion of implications for future user-centered attack-aware systems. This work can serve as a basis for usable security researchers to better design systems against user-centered attacks.
Conference Paper
Full-text available
Personal mobile devices keep private information which people other than the owner may try to access. Thus far, it has been unclear how common it is for people to snoop on one another's devices. Through an anonymity-preserving survey experiment, we quantify the pervasiveness of snooping attacks, defined as "look-ing through someone else's phone without their permission." We estimated the 1-year prevalence to be 31% in an online participant pool. Weighted to the U.S. population, the data indicates that 1 in 5 adults snooped on at least one other person's phone, just in the year before the survey was conducted. We found snooping attacks to be especially prevalent among young people, and among those who are themselves smartphone users. In a follow-up study, we found that, among smartphone users, depth of adoption, like age, also predicts the probability of engaging in snooping attacks. In particular, the more people use their devices for personal purposes, the more likely they are to snoop on others, possibly because they become aware of the sensitive information that is kept, and how to access it. These findings suggest that, all else remaining equal, the prevalence of snooping attacks may grow, as more people adopt smartphones, and motivate further effort into improving defenses.
Full-text available
Drones are unmanned aircraft controlled remotely or operated autonomously. While the extant literature suggests that drones can in principle invade people’s privacy, little is known about how people actually think about drones. Drawing from a series of in-depth interviews conducted in the United States, we provide a novel and rich account of people’s privacy perceptions of drones for civilian uses both in general and under specific usage scenarios. Our informants raised both physical and information privacy issues against government, organization and individual use of drones. Informants’ reasoning about the acceptance of drone use was in part based on whether the drone is operating in a public or private space. However, our informants differed significantly in their definitions of public and private spaces. While our informants’ privacy concerns such as surveillance, data collection and sharing have been raised for other tracking technologies such as camera phones and closed-circuit television (CCTV), our interviews highlight two heightened issues of drones: (1) powerful yet inconspicuous data collection, (2) hidden and inaccessible drone controllers. These two aspects of drones render some of people’s existing privacy practices futile (e.g., notice recording and ask controllers to stop or delete the recording). Some informants demanded notifications of drones near them and expected drone controllers asking for their explicit permissions before recording. We discuss implications for future privacy-enhancing drone designs.
Conference Paper
Full-text available
Conference Paper
We present a novel concept for protecting text messages (e.g. notifications) on mobile devices from shoulder surfing. We propose to display the text in the user's handwriting, assuming that people can read their own handwriting easier and faster than strangers. Our approach was evaluated in a proof-of-concept user study that revealed significant differences in reading time: Participants were indeed slower when reading the unfamiliar handwriting of the other participants compared to their own, and they tended to make more errors. Even though this effect was not present for all participants, we argue that our results may provide the basis for protection mechanisms applicable in real-world scenarios.
Conference Paper
We present an approach to protect photos on smartphones from unwanted observations by distorting them in a way that makes it hard or impossible to recognize their content for an onlooker who does not know the photographs. On the other hand, due to the chosen way of distortion, the device owners who know the original images have no problems recognizing photos. We report the results of a user study (n=18) that showed very high usability properties for all tested graphical filters (only 11 out of 216 distorted photos were not correctly identified by their owners). At the same time, two of the filters significantly reduced the observability of the image contents.
Conference Paper
While media reports about wearable cameras have focused on the privacy concerns of bystanders, the perspectives of the `lifeloggers' themselves have not been adequately studied. We report on additional analysis of our previous in-situ lifelogging study in which 36 participants wore a camera for a week and then reviewed the images to specify privacy and sharing preferences. In this Note, we analyze the photos themselves, seeking to understand what makes a photo private, what participants said about their images, and what we can learn about privacy in this new and very different context where photos are captured automatically by one's wearable camera. We find that these devices record many moments that may not be captured by traditional (deliberate) photography, with camera owners concerned about impression management and protecting private information of both themselves and bystanders.
Conference Paper
In this paper, we present SwiPIN, a novel authentication system that allows input of traditional PINs using simple touch gestures like up or down and makes it secure against human observers. We present two user studies which evaluated different designs of SwiPIN and compared it against traditional PIN. The results show that SwiPIN performs adequately fast (3.7 s) to serve as an alternative input method for risky situations. Furthermore, SwiPIN is easy to use, significantly more secure against shoulder surfing attacks and switching between PIN and SwiPIN feels natural.