You are on page 1of 9

Computers in Human Behavior 26 (2010) 1739–1747

Contents lists available at ScienceDirect

Computers in Human Behavior


journal homepage: www.elsevier.com/locate/comphumbeh

It won’t happen to me: Promoting secure behaviour among internet users


Nicola Davinson *, Elizabeth Sillence
PaCTLab, School of Psychology and Sports Science, Northumbria University, Newcastle upon Tyne NE1 8ST, UK

a r t i c l e i n f o a b s t r a c t

Article history: Fraudulent activity on the Internet, in particular the practice known as ‘Phishing’, is on the increase.
Available online 31 July 2010 Although a number of technology focussed counter measures have been explored user behaviour remains
fundamental to increased online security. Encouraging users to engage in secure online behaviour is dif-
Keywords: ficult with a number of different barriers to change. Guided by a model adapted from health psychology
Security this paper reports on a study designed to encourage secure behaviour online. The study aimed to inves-
Risk perception tigate the effects of education via a training program and the effects of risk level manipulation on subse-
Risk information
quent self-reported behaviour online. The training program ‘Anti-Phishing Phil’ informed users of the
Training
Phishing
common types of phishing threats and how to identify them whilst the risk level manipulation randomly
Internet allocated participants to either high risk or low risk of becoming a victim of online fraud. Sixty-four par-
ticipants took part in the study, which comprised of 9 males and 55 females with an age range of 18–
43 years. Participants were randomly allocated to one of four experimental groups. High threat informa-
tion and/or the provision of phishing education were expected to increase self-reports of secure behav-
iour. Secure behaviour was measured at three stages, a baseline measure stage, an intention measure
stage, and a 7-day follow-up measure stage. The results showed that offering a seemingly tailored risk
message increased users’ intentions to act in a secure manner online regardless of whether the risk mes-
sage indicated they were at high or low risk of fraud. There was no effect of the training programme on
secure behaviour in general. The findings are discussed in relation to the model of behaviour change,
information provision and the transferability of training.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction & Odabasi, 2007), for example by providing further opportunity


for academic dishonesty among students (Akbulut et al., 2008),
The number of people using the Internet is constantly increas- and providing additional channels for aggressive or threatening
ing; with a US survey conducted in 2006 indicating 73% of their behaviour, more recently known as ‘‘cybervictimisation” (Akbulut,
respondents use the Internet, whereas the same survey carried Sahin, & Eristi, 2010). More specifically in a financial domain, the
out in 2005 reported 66% were Internet users. Additionally, statis- practice known as ‘Phishing’ is on the increase. Myers (2007, p. 1)
tics reported by Internet World Stats (August 2008) indicate the describes phishing as ‘‘a form of social engineering in which an at-
number of Internet users worldwide has increased 305.5% since tacker, also known as a ‘phisher’, attempts to fraudulently retrieve
2000, now representing 21.9% of the population worldwide. The legitimate users’ confidential or sensitive credentials by mimicking
number of people using the Internet to perform financial transac- electronic communications from a trustworthy or public organisa-
tions has similarly increased. It is estimated that 2012 will see tion in an automated fashion.” In December 2009, there were
one billion users completing online purchases (IDC, June 2008). 46,190 unique phishing sites detected (Anti-Phishing Working
The Internet offers a number of benefits to consumers, such as Group, APWG) and in recent years phishing methods have become
24 h accessibility (Berthon, Pitt, & Watson, 1996), improved interac- much more prevalent and sophisticated (Milletary, 2005; Moore &
tivity (Maignan & Lukas, 1997), and convenience (Jarvenpaa & Todd, Clayton, 2007; Myers, 2007). Typically victims receive an email dis-
1997; Wolhandler, 1999). Accessibility and functionality improve- guised as an authentic organisation which asks them to re-confirm
ments such as these have made it quicker and easier for users to their details by following a link. The link directs them to a spoof
purchase goods and services online and to manage their bank ac- website that often looks exactly the same or at the very least uses
counts via the Internet. However, these advancements also make the same images and logos as the genuine site. The proprietor of
unethical use of the Internet easier and more widespread (Namlu the spoofed website then harvests the information given by the vic-
tim, typically bank details, and uses it fraudulently. The costs of
* Corresponding author. Tel.: +44 191 227 3716; fax: +44 191 227 3190. phishing are not simply the direct losses associated with the crime,
E-mail address: nicola.davinson@northumbria.ac.uk (N. Davinson). but also the indirect and opportunity costs that are incurred (Myers,

0747-5632/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.chb.2010.06.023
1740 N. Davinson, E. Sillence / Computers in Human Behavior 26 (2010) 1739–1747

2007). These include the costs associated with devalued brand and sion as a way to con them into supplying their username and pass-
the loss of actual and potential online customers. word information or even their bank details. As noted previously
From a technology perspective there are a number of counter- this type of phishing attack is often difficult for users to detect ren-
measures to phishing available (Emigh, 2007). These include email dering credibility evaluation difficult.
filtering and email authentication which would prevent the user
from actually receiving the phishing attack in the first instance.
1.2. Credibility evaluation
There are also a number of other technological security systems
that can be put in place, such as the introduction of phishing tool-
We know that users consider a number of factors when assess-
bars and indicators that make it easier for the user to identity a
ing the credibility of a website and that these factors differ from
fraudulent site. Typically, these kinds of security measures are only
expert evaluations (Stanford, Tauber, Fogg, & Marable, 2002). Large
concerned with the technological side of the system and fail to
scale surveys by Fogg et al. (2001, 2003) suggest that credibility is
consider the user. Users, however, could be seen to be the ‘weakest
primarily driven by an attractive and professional design, or is
link in the security chain’ (Schneier, 2000) suggesting the need for
influenced by the presence or absence of visual anchors or promi-
a greater emphasis on user awareness and education in terms of
nent features such as a photograph or trust seal (Riegelsberger,
online security.
Sasse, & McCarthy, 2003). Others suggest that credibility evalua-
This paper reports on a study which examines whether users
tions are based upon judgments of the institution, degree of
can be encouraged to protect themselves through a combination
personalisation, communication integrity or like-mindedness
of increased awareness and training enabling them to adopt a more
(Bhattacherjee, 2002; Briggs, de Angeli, & Simpson, 2004; Egger,
secure style of interaction when conducting financial transactions
2000; Sillence, Briggs, Harris, & Fishwick, 2007). Developmental
online. Before presenting the study itself it is worth reviewing
or staged models of credibility (or a related construct trust) help
the literature surrounding user awareness and behaviour in this
reconcile these different emphases. Staged models (e.g. Sillence,
domain as well as introducing a potentially useful model for guid-
Briggs, Harris, & Fishwick, 2006) also allow us to distinguish be-
ing our thinking about behavioural change.
tween relatively hasty and more considered processing strategies
for evaluating credibility online. Their model suggests that users
1.1. User concerns and behaviour
first of all carry out a rapid screening of the sites based upon a heu-
ristic analysis. During this stage they base their credibility deci-
It is apparent that consumers have concerns about the security
sions upon predominantly visual features, e.g. colour, logos, and
and privacy of the data they may provide to websites (Briones,
a familiar layout. Given sufficient time and resources, e.g. knowl-
1998; Culnan, 1999). Internet users report being concerned about
edge and motivation, the user will then engage in a more system-
the collection and use of the personal information they supply
atic analysis of the website’s content. These resources are,
when shopping online (Rohm & Milne, 1998; Sheehan & Hoy,
however, often absent and Dhamija, Tygar, and Hearst (2006)
2000). Such concerns have previously been thought to be only a
noted that people often lack the specific knowledge to spot the vi-
minor barrier to the adoption of online shopping (Jarvenpaa &
sual deception present in a phishing scam. To complicate matters
Todd, 1996–1997); however, more recently it is thought that such
Emigh (2007) notes that organisations, such as American Express
concerns play a major role in consumers’ willingness to purchase
and Capital One, include one or more of the common errors that
goods online (Culnan, 1999; Metzger, 2004). Previous research,
users are warned to look out for, e.g. do not click on email links,
for example, has indicated that the primary reason young consum-
check for the padlock icon and check for a valid SSL session before
ers choose not to conduct financial transactions online is due to
entering data. It seems likely that faced with a vast array of infor-
concerns about the privacy of their credit card information
mation in the form of web search results or spam emails, the aver-
(McQuivey & Ham, 2000). Similarly, Fram and Grady (1997) report
age user will fall back on heuristic credibility evaluation processes.
a study relating to security aspects of online purchasing and state
when a consumer has concerns about credit card fraud they will
only commit to purchases with a low level of risk. Furthermore, 2. Modifying user behaviour
risk perception does not only affect online purchases, but has also
been found to affect adoption of Internet banking (Tan & Teo, Thus we have seen that credibility evaluation is not straightfor-
2000). ward. Users are often faced with time pressures or insufficient
However, this apparent concern among Internet users is often knowledge or motivation to carry out anything other than a heuris-
not noticeable in their online actions when completing financial tic analysis of online material. This is just one of the barriers facing
transactions. Studies have indicated that users are willing to users that researchers need to consider when trying to decide how
trade-off their privacy concerns in return for benefits such as con- to encourage users to behave more securely online. Users may not
venience (Chellappa & Sin, 2005; Spiekermann, Grossklags, & Ber- always weigh up the costs and benefits associated with good secu-
endt, 2002). Therefore, it has been suggested that there is a rity practices accurately, for example believing financial institu-
dichotomy between attitudes and behaviour in this domain. Spie- tions to be solely responsible for preventing fraud (Davinson &
kermann et al. (2002) found that participants who had been classi- Sillence, in preparation). How then can we increase user motiva-
fied as privacy advocates were still willing to trade their personal tion to engage in secure behaviour online? Whilst there is rela-
information to obtain a small benefit. This dichotomy is by no tively little literature regarding behavioural change in financial
means a new phenomenon and has been present and detected in settings; a useful starting point is the health psychology literature
other domains, such as attitudes and behaviour towards racial pre- which provides a number of behaviour based models. The Health
judice and classroom cheating (Corey, 1937; LaPiere, 1934), indeed Belief Model (HBM) (Becker & Rosenstock, 1987; Rosenstock,
there is a wealth of literature which considers the cause and effect 1966) predicts that behaviour is due to a number of user percep-
nature of attitudes and behaviour (Ajzen, 1985; Festinger, 1957). tions concerning threat, costs and benefits and control. Table 1 dis-
If users are willing to trade-off their privacy attitudes in order to plays the elements of the HBM and then how they could
receive benefits it is likely that they will also behave against their potentially be adapted for a financial context.
privacy judgements in order to avoid costs, for example, having Practicing secure behaviour is expected to be affected by per-
their online accounts suspended. Many online fraudsters prey on ception of a threat being present and whether enacting a secure
user insecurities and use the threat of account closure or suspen- behaviour will actually reduce the threat. Perception of a threat
N. Davinson, E. Sillence / Computers in Human Behavior 26 (2010) 1739–1747 1741

Table 1 place. Also, some factors such as motivation are less tangible and
Health belief model and potential financial considerations. less directly observable and appear to be a product of other factors
Concept Financial context (perceived susceptibility and severity). Therefore, perceived sus-
Perceived What are the chances of becoming a victim of fraud? ceptibility will be focussed upon initially, as it is reasonable to as-
susceptibility sume this factor will increase concerns and motivation to act,
Perceived severity How serious fraud and its consequences are? which must be satisfied before a change in behaviour is even
Perceived costs What are the costs involved in carrying out the secure considered.
behaviour?
Perceived benefits What are the benefits of carrying out secure behaviour?
Cues to action When to conduct secure behaviour?
Health motivation Am I concerned that technology use can affect my 2.1.1. Increasing users’ motivation via perceived susceptibility
finances? One way of increasing motivation is to raise the level of user
Perceived control Can I prevent fraud by behaving securely? knowledge in the domain. Providing users with more information
about the threat of online fraud could increase their perceptions
of susceptibility. Studies have investigated the effective use of
is likely to be affected by susceptibility, severity, and motivation, threat appeals for health communication (Leventhal, 1970; Rogers,
and evaluation of secure behaviour is likely to be affected by costs, 1975) and have attempted to utilize threat in changing health be-
benefits, and control. Once a threat is perceived and secure behav- liefs (Kirscht & Haefner, 1973; Radelfinger, 1965). For example,
iour is chosen the user must also know when to conduct it. There is Becker, Maiman, Kirscht, Haefner, and Drachman (1977) indicate
little evidence to indicate a predetermined sequence of events that that the use of fear-arousal did have an effect on recorded weight
must occur to successfully promote secure behaviour, or which loss, with the high fear intervention being the most consistent in
factors need to be satisfied before others come into play; however, the long-term. The low fear intervention also recorded more
it is logical to suggest a sequence in the order presented from left weight loss than the control group but appears to be less effective
to right in Fig. 1. than the high fear group. It is also noted that the experimental
Reasonable order of events treatments were most effective for those who had the lowest be-
Firstly, if users are not informed about technology breaches and liefs about threat initially.
risk when completing financial transactions they will not perceive Whilst research throughout the 60s and 70s into the value of
susceptibility to be high. Similarly, if users do not consider the con- using threat appeals is somewhat inconclusive (McGuire, 1968)
sequences of fraud to impact their lives or their finances they will there is some suggestion that the evidence favours the use of threat
not perceive severity of fraud to be high. If susceptibility to the when making health recommendations (Leventhal, 1973). It is also
threat and severity of the consequences are not perceived the user noted that the threat itself is not necessarily the direct link to
will not be concerned about protecting their finances and ulti- behaviour change but that the fear produced from such threats is
mately will not be motivated to behave securely when conducting something that people wish to resolve, and hence indirectly threat
financial transactions. However, even if users are sufficiently moti- should affect behaviour (e.g. Janis, 1974). Leventhal (1973) points
vated, feel susceptible, and perceive severe consequences it does out that the use of threat is particularly useful when provided
not necessarily change their behaviour. Practices are also governed alongside methods of coping with the threat, especially among
by perceived control, perceived costs and benefits, and finally cues those with low levels of concern.
to action. The user must perceive they can control the occurrence It is also important to consider coping strategies when users are
of fraud, for instance checking for the relevant security symbols, presented with threat information. Rippetoe and Rogers (1987)
such as the lock symbol, when making purchases to ensure details investigated the impact of threat (defined as severity plus vulner-
are being entered into a secure site will help to prevent fraud ability), response efficacy and self-efficacy on two adaptive coping
occurring. These symbols, however, can also be spoofed by phish- responses and five maladaptive coping responses. Findings indi-
ers therefore rendering the action redundant. Furthermore, the cate that a high level of threat encourages the individual to act
user must also decide if the potential benefits outweigh the costs, regardless of whether the response was adaptive or maladaptive.
for instance checking for the lock symbol requires minimal effort In addition, the coping strategy chosen by participants was depen-
and could potentially prevent fraudulent use of their information. dent upon which coping appraisal information they had been gi-
Nevertheless, if all three of the factors are satisfied the user must ven. Specifically, the effective coping information (high response
then know when to carry out the correct preventative practices efficacy and high self-efficacy) led to greater use of the adaptive
otherwise secure behaviour may not be adequately achieved. The rather than maladaptive strategies when compared to the ineffec-
current focus for this research is on identifying and examining tive coping information. Therefore, caution must be taken when
the factors in relation to promoting secure behaviour rather than high threat is present in the absence of effective coping strategies
specifying the precise sequence of the events. as it appears participants are likely to choose a maladaptive strat-
egy in response to high threat if an adaptive strategy is not known
or available.
2.1. Promoting secure behaviour Thus, once motivated via increased susceptibility the user also
needs to be aware of an effective coping strategy to prevent the
It appears logical that the effect of some factors is engaged be- occurrence of fraud. In the Internet security and phishing domain
fore others, for example it is unlikely that a user will need to over- this typically entails having the knowledge and means to be able
come perceived costs if they are not motivated to act in the first to spot phishing websites and avoid using them. Previously there

Fig. 1. Hierarchy of factors contributing towards secure behaviour.


1742 N. Davinson, E. Sillence / Computers in Human Behavior 26 (2010) 1739–1747

has been a reliance on technology to aid the user in identifying erated to indicate ‘high’ or ‘low’ risk and did not reflect how they
phishing sites, for example through the use of toolbars (e.g. had responded to the questionnaire; however, the participants
www.spoofstick.com). Some researchers, however, are exploring were led to believe the score was based on their questionnaire re-
ways of training users to reduce their risk of suffering phishing at- sponses. Secondly, participants were then asked to complete a
tacks. A group at Carnegie Mellon University have manipulated questionnaire to measure their intentions to behave securely over
user education and training noting the success of embedded train- the next 7 days. Thirdly, participants in the training group were
ing materials in which users are ‘‘learning by doing” in relation to then asked to complete the Anti-Phishing Phil education program.
phishing attacks (Kumaraguru et al., 2007). Users that receive Finally, all participants were emailed 1 week later to measure their
training after they have fallen for a phishing scam are more likely secure behaviour in the 7 days since they had completed the study.
to retain and transfer their knowledge than those who simply
received the training materials via email. In a similar vein, 3.1. Participants
Sheng et al. (2007) successfully demonstrated the usefulness of
interactive learning experiences in increasing users’ ability to spot Sixty-four participants completed the study in full which com-
phishing sites. They developed an interactive training program prised of 9 males and 55 females (total average age 21.97 years)
‘Anti-Phishing Phil’ to help users identify and prevent phishing with an age range of 18–43 years. It is noted that the sample is pre-
attacks. In this game the user plays throughout four rounds and dominantly female which may be somewhat of a drawback, how-
is given tutorials between each round as the attacks become ever, the proportion of males/females recruited does reflect the
increasingly difficult to spot. Sheng et al. (2007) found that playing Psychology student population from which they were sought. Par-
the interactive game did result in greater identification of phishing ticipants were randomly allocated using a random number gener-
sites among users compared with users who viewed current online ator to one of four experimental groups.
tutorials or a paper based version of the interactive game. Behav-
ioural studies using role play scenarios have also noted the impor-  Group 1: Low threat/no training consisted of 2 males and 14
tance of education in helping people to avoid phishing attacks females (average age 21.2 years). One participant was a previ-
(Downs, Holbrook, & Cranor, 2007). ous victim of online fraud.
Therefore, the current study proposes to raise levels of user  Group 2: Low threat/training consisted of 1 male and 15
motivation through an increase in perceived susceptibility. This females (average age 21.98 years). One participant was a previ-
will be achieved by providing information about the methods ous victim of online fraud.
and consequences of phishing and online fraud alongside a tailored  Group 3: High threat/no training consisted of 3 males and 13
risk warning score. The information given to participants will con- females (average age 22.13 years). One participant was a previ-
tain both reassuring and threatening elements as the aim is to in- ous victim of online fraud.
crease feelings of threat but also to reassure the user that they can  Group 4: High threat/training consisted of 3 males and 13
do something to prevent the threat. It is expected that those in the females (average age 22.63 years). Two participants were previ-
high risk warning group will report more secure behaviour inten- ously victims of online fraud.
tions and more secure behaviour at a 1-week follow-up. A further
aim of the study is to provide users with an effective coping strat- It is noted that 8% of participants (n = 5) reported previously
egy to ensure the increase to perceived susceptibility is most effec- being victims of online fraud and 42% (n = 27) of participants re-
tive in motivating adaptive responses and changes in behaviour. It ported knowing someone who had previously been a victim of on-
is expected that educating users by means of the interactive game line fraud.
‘Anti-Phishing Phil’ will provide users with an effective coping Participants were asked a number of demographic questions
strategy and will also result in more secure behaviour reports, spe- relating to their Internet use, including length of use of the Internet,
cifically at the follow-up stage. Therefore, this study hypothesises shopping online and banking online. The majority of participants
users will report a greater perception of susceptibility and will re- have used the Internet for over 3 years, have been shopping online
port more secure behaviour when given a high risk warning score for longer than 12 months, and banking online for less than
and/or are provided with training to overcome the threat. 12 months. Most participants use the Internet for more than 5 h/
week, with 50% indicating the used the Internet in excess of 10 h/
3. Method week. Similarly most participants had made at least one purchase
online in the past month, with 35% indicating they had made more
The study consisted of four main stages (Fig. 2). Firstly, partici- than five purchases online in the same period. Finally, there was a
pants were presented with an online questionnaire to determine a very similar number of people who reported using Internet banking
baseline score of their online behaviour which was retained for monthly (34%) weekly (34%) or not at all (28%).
comparison purposes with post-manipulation and 1-week fol-
low-up. Upon completion of the questionnaire a risk score was ran- 3.1.1. Randomisation
domly generated together with appropriate risk information to ANOVA analysis indicates no significant differences between
inform the participants whether they were at high risk or low risk the groups for their baseline score of behaviour. Furthermore, v2
of becoming a victim of online fraud. This score was randomly gen- analyses for independence were conducted on the demographic

Fig. 2. Methodology sequence of procedure.


N. Davinson, E. Sillence / Computers in Human Behavior 26 (2010) 1739–1747 1743

data per group to ensure random allocation had occurred. There details retrieved from getsafeonline.org and the APACS website.
were no significant relationships between any of the demographic The only information which differed per group included a preced-
variables and the groups except for experience of online banking ing paragraph which is presented below:
(v2(6) = 13.145, p = 0.041). The high threat score/training given
‘‘Well done! You have a low percentage risk of becoming a victim of
group appear to have more experience of banking online than
fraud due to the way you use the Internet. You appear to be acting
the other groups with 10% reporting more than 3 years of experi-
safely while using the Internet however do not become complacent
ence compared to 3% and less reported in the other groups.
as threats are evolving all the time.”
Participants were recruited from the population of staff and stu-
‘‘Warning! You have a very high percentage risk of becoming a vic-
dents at Northumbria University with the stipulation that they had
tim of fraud due to the way you use the Internet.”
previous experience of using the Internet for purchases and/or
banking. The study was advertised via the Division of Psychology
notice board and participants were also sought through canvassing 3.3.5. Training program
of students during seminars. Incentives were provided in the form Participants in the training group were asked to complete a
of payment and course credit. Ethical approval was sought and training program ‘Anti-Phishing Phil’ (Fig. 4) which was developed
granted from the Psychology and Sport Science School Ethics Com- by, and used with permission of the CUPS lab (CMU Usable Privacy
mittee at Northumbria University. and Security) at Carnegie Melon University, USA. The program is a
fun and interactive game that trains the user to be more able to de-
3.2. Design fend themselves against phishing attacks. The program consists of
four rounds with tutorial information before each round. Each
The study had an experimental design, with two between par- round requires the participant to identify the credibility of URLs
ticipants independent variables: participants were randomly as- by determining whether they are genuine or not, with the tutorial
signed to one of four groups (1) low risk warning score/no information providing information on the appropriate cues to look
training (n = 16); (2) low risk warning score/training given out for when assessing the credibility of web addresses. Partici-
(n = 16); (3) high risk warning score/no training (n = 16); (4) high pants were asked to complete all four rounds and were given as
risk warning score/training given (n = 16). The experiment had a much time as they required, typically taking 15–20 min.
prospective component, with a follow-up after 1 week.
3.4. Procedure
3.3. Materials
Participants were tested individually and upon arriving at the
3.3.1. Baseline measure laboratory they were told the study involved an evaluation of risk
Measures were taken of age, sex, Internet use, experience of car- when conducting financial transactions online. After completing
rying out online financial transactions and experience of fraud. the baseline measures, participants were asked to wait for their on-
Baseline or current behaviour was measured using an 11 item scale line risk warning score to be calculated. They were asked to read
(a = 0.829), e.g. ‘‘I only use websites with the secure padlock icon the information provided to them regarding their score before
when shopping online” and ‘‘I only use reputable companies when completing the second paper-based questionnaire. The training
shopping online” and utilised a seven-point Likert scale (1 = Al- groups were then directed to the Anti-Phishing Phil training pro-
ways – 7 = Never). Susceptibility was measured using two items gram and asked to spend as much time as they required complet-
(a = 0.734) and utilised a seven-point Likert scale (1 = Strongly dis- ing all four rounds. Instructions on how to play the training game
agree – 7 = Strongly agree). All measures were taken via an online were presented as a part of the training program. Participants were
questionnaire delivered by Survey Monkey. sent the follow-up measures by email approximately 7 days later
at which point they were fully debriefed.
3.3.2. Intention measure
A further paper-based questionnaire was administered consist-
ing of the same behaviour items and the same susceptibility items 4. Results
that were used for the baseline measure but framed to reflect
intention over the next 7 days. For example: In the next 7 days I in- 4.1. Effects of risk warning score and training
tend to ‘‘only use websites with the secure padlock icon when
shopping online” and responses were recorded on a seven-point The sum of the behaviour items at each stage of the study was
Likert scale (1 = More often – 7 = Less often). calculated with a lower score indicating more security-conscious
and responsible behaviour. Data was analysed using a 2  2  3
3.3.3. One-week follow-up measure (Risk warning score  Training  Stage) mixed design ANOVA,
After 1 week participants received a brief follow-up question- with Stage as the only within factor. There was a significant main
naire, by email. Included were the same behaviour and susceptibil- effect of stage of questionnaire completion (F(2,120) = 43.837,
ity items but framed to reflect the previous 7 days. For example: In p < 0.001) and pairwise comparisons indicate significant differ-
the past week, I have ‘‘only used reputable companies when shop- ences (p < 0.001 and p = 0.003) between all three stages (Fig. 3).
ping online” and responses were recorded on a seven-point Likert Therefore, it appears once given the risk warning score, regardless
scale (1 = More often – 7 = Less often). of whether it indicated high or low risk, participants intentions
were to behave more securely than they had indicated at baseline.
3.3.4. Risk warning score presentation In contrast at the follow-up stage, regardless of risk warning score
Following completion of the baseline measure participants or training, their reported behaviour was significantly less secure
were told that their ‘risk warning score’ was being calculated. This than they had intended to be. However, it was still significantly
was then presented to the participants. Participants in the low risk more secure compared to the initial baseline measure.
warning score groups were presented with a ‘‘20% at RISK” fact There was no main effect of risk warning score or training at any
sheet, whereas participants in the high risk warning score groups of the three stages, suggesting that users only need to be given
were presented with an ‘‘80% at RISK” fact sheet. Each group in fact information at a generic level and do not need to have their level
received exactly the same information which was a combination of of risk to be increased in order to behave more securely.
1744 N. Davinson, E. Sillence / Computers in Human Behavior 26 (2010) 1739–1747

ring to another person increases significantly. Paired samples t-


tests also indicated that at each stage participants consistently per-
ceived there to be more risk to others than to themselves.

5. Discussion and conclusions

The aim of this study was to encourage users to behave more


securely when using the Internet to complete financial transac-
tions through a combination of increased perception of susceptibil-
ity and identifying a coping strategy. It was expected that a greater
perception of susceptibility would motivate the user to think more
carefully about whom they gave their personal details to, and in
what circumstances they would disclose information, when they
use the Internet. Similarly, it was expected that providing educa-
Fig. 3. Reported secure behaviour at each of the three stages, with a lower score
tion through an interactive training programme would provide
indicating more secure behaviour reported.
the user with an effective coping strategy, and that in turn this
would help the user to understand how they can combat phishing
4.2. Susceptibility perceptions
Table 2
The perceived susceptibility items included perceived suscepti- Mean score pertaining to likelihood of becoming a victim of fraud, with a higher score
bility to the individual and perceived susceptibility to someone indicating a greater likelihood.
else and there were significant differences between the two items.
Baseline Intention Follow-up
With regard to themselves there was a significant affect of stage of measure measure measure
completion (F(1,18) = 10.068, p < 0.001). Pairwise comparisons indi-
Mean perceived susceptibility 3.560 4.233 3.900
cate only one significant difference (p < 0.001) between the base- to self-score
line measure and the intention measure (Table 2). Therefore, it
appears that following the risk warning score, regardless of
whether it indicated high or low risk, perceived susceptibility to
the occurrence of online fraud increased. Table 3
Mean score pertaining to the likelihood of someone else becoming a victim of online
With regard to fraud occurring to another person there was also
fraud, with a higher score indicating a greater likelihood.
a significant affect of stage (F(1,118) = 4.641, p = 0.011) with pairwise
comparisons indicating a significant difference between the base- Baseline Intentions Follow-up
measure measure measure
line measure and the intention measure (Table 3). Again it appears
that following the score information, regardless of whether it in- Mean perceived susceptibility 4.441 4.888 4.858
to others score
cluded a high or low risk, the perceived likelihood of fraud occur-

Fig. 4. Screenshot taken from round 2 of the Anti-Phishing Phil training program.
N. Davinson, E. Sillence / Computers in Human Behavior 26 (2010) 1739–1747 1745

attacks that rely on disclosure of personal information. Whilst cated than first thought. A study by LaRose, Rifon, and Enbody
users did report behaving more securely, the data present a more (2008) may provide an alternative explanation for this finding.
complex picture concerning perceived susceptibility and training. They found that moderate levels of threat susceptibility were least
The findings are discussed below in terms of both the practical related to safe behaviours, such as updating security patches and
and theoretical implications. scanning for spyware, and those with high or low threat suscepti-
Following the presentation of the risk warning score users re- bility were more likely to behave in a secure manner. Thus, we can
ported that they intended to behave more securely (compared to be somewhat confident that the findings relating to intentions
the baseline behaviour) and reported an increased likelihood of were not only due to the ‘prompt’ provided by the warning infor-
becoming a victim of fraud (compared to the baseline measure). mation, although the effect of such a prompt cannot be ruled out
Similarly, the follow-up stage indicated an increase in secure entirely.
behaviour compared to baseline measures, which was regardless This result, however, also conflicts with previous research that
of both risk and training level. This shows that the risk information suggests interventions that simply inform people they are suscep-
was successful to some extent in raising perceived susceptibility to tible to health risks are often not sufficient to change their behav-
motivate the user into behaving in a more secure way. Therefore, it iour (Leventhal et al., 1997). In fact there is an education tool in
suggests that giving risk information to Internet users will encour- the health domain which has been criticised because it simply re-
age them to behave more securely when they complete financial lies on perceptions of risk to change behaviour. The Health Risk
transactions online. The fact that the two different risk warning Appraisal (HRA) relies on risk perception alone to be effective in
scores failed to differentiate the participants’ susceptibility feelings changing behaviour and does not provide sufficient information
may indicate that the materials used were underpowered to some regarding how to make the changes they suggest. Kreuter and
extent. From an ethical standpoint developing materials which Strecher (1996) attempted to enhance the HRA technique by
would instil a greater sense of susceptibility is difficult. The fol- including the assessment of relevant psychological factors, such
low-up aspect to the study meant that participants would be left, as costs and benefits, and facilitating self-change via individua-
potentially anxious, for 7 days before being debriefed. lised feedback. Their findings suggest a positive effect of individu-
Therefore, it was necessary to ensure that the information was ally tailored messages to the HRA tool in promoting change in
both reassuring and threatening, so that the user was not unduly three of the seven health related areas under investigation. When
anxious. This may have resulted in the two risk warning messages tailored messages were not received the HRA was not effective in
being too similar. any of the areas. This may also help to explain why the current
The training programme utilised in the study to provide an study elicited a change in behaviour via raised awareness of sus-
effective coping strategy has previously been shown to increase ceptibility as users were led to believe that they were receiving
the ability of users to detect phishing websites (Sheng et al., a tailored risk warning based on their current behaviour. There
2007). It was hoped that the training would raise participants’ is evidence to suggest tailored messages are more likely receive
knowledge at a specific and general level. Both with regard to attention and be recalled (Campbell et al., 1994) and previous re-
the actual websites featured in the programme and to the broader search indicates a positive effect of tailoring health information
principles of behaving securely online. Training programmes in particularly in the areas of smoking, diet and mammography
phishing studies have shown positive ‘near’ (Whitten & Bjork, attendance (Campbell et al., 1994; Skinner, Strecher, & Hospers,
1977) transfer results, i.e., a transfer of knowledge when the train- 1994; Strecher et al., 1994).
ing and testing materials were very similar (Kumaraguru et al., It is also worth considering the possible role of the self-proph-
2007). In this study, the training appears not to have had a broad ecy effect (Sprott, Spangenberg, & Fisher, 2003) on the findings re-
effect beyond the very specific points it trained on. The current ported here. This suggests simply asking users whether they intend
study was concerned with promoting secure behaviour in a real to undertake a particular behaviour is enough to increase the prob-
world setting. The training programme used was perhaps too spe- ability that they will do so. This effect has been shown across a
cific for the range of Internet experiences tested out by the partic- number of domains, such as voting behaviour (Greenwald, Carnot,
ipants. It may be that more thought needs to be given to Beach, & Young, 1987). It also indicates that simply making the
developing a broader security education training programme. user aware and providing the opportunity to think about secure
behaviour can be successful in changing their subsequent behav-
5.1. Raising awareness of susceptibility iour. Furthermore, providing users with education and indicating
effective coping strategies for them to undertake is certainly not
The fact that secure behaviour increased at both the intention harmful to this effect, and it is unlikely that users will sustain se-
stage and at the 7-day follow-up regardless of information type cure behaviour over a longer period if they are unaware of the con-
or training is of particular interest. It may be that simply providing sequences or coping strategies available.
the risk warning information raised awareness of the risks suffi- There are a number of practical applications for the findings of
ciently to encourage intention and behavioural change. The fact this study, firstly in the contribution to future interventions aimed
that it is either ‘highly likely to happen’ or ‘quite unlikely to hap- at increasing secure behaviour when using technology to complete
pen’ may almost be a side issue. For some people simply being re- financial transactions. For example, the findings of this study indi-
minded that the risk exists may be enough to promote intention cate that fear appeals can be used effectively in this domain which
and behaviour change. The findings relating to intentions to be- is strengthened by personalisation of the message. Secondly, there
have may appear unsurprising given the warning information are wider implications for providers of e-commerce and e-banking
which was presented immediately prior to the measure being ta- domains to understand consumer issues surrounding fraudulent
ken. However, participants were presented with different levels attacks. For example, effectively educating the user to identify
of warning and a meta-analysis of fear appeals suggests strong fear appropriate cues which indicate credibility will not only help to
appeals are more persuasive than low or weak fear appeals (Witte decrease successful fraudulent attacks but will also increase user
& Allen, 2000). Therefore, it would be expected that those in the satisfaction and uptake of services online if they are confident in
high threat condition would feel more compelled to change their their abilities to use the services safely.
behaviour than those in the low threat condition. The fact that In theoretical terms, this study has provided some support for
all participants reported intentions to behave more securely, the use of the adapted health belief model as a way of predicting
regardless of the condition, indicates the finding is more compli- secure behaviour in a financial context.
1746 N. Davinson, E. Sillence / Computers in Human Behavior 26 (2010) 1739–1747

Raising awareness of susceptibility does appear to increase se- plenty of opportunity to practice the secure behaviour which they
cure behaviour. A useful next step would be to investigate suscep- reported.
tibility alongside severity. Although Downs et al. (2007) found that Finally, it is apparent that the follow-up behaviour reported is
severity per se did not seem to encourage secure behaviour it may less secure than the reported intention behaviour, which raises
be that combinations of the two factors can prove effective in pro- questions regarding the longevity of such an intervention. Future
moting change. The findings also reflect the intention-behaviour research would benefit from an extended follow-up period that
gap common in much of the social and health behaviour literature can assess just how long the message lasts. An extended follow-
(e.g. Sheeran, 2002). Intentions to behave securely were signifi- up could also serve to provide more depth to the current findings,
cantly higher than the actual behaviour reported 7 days later. perhaps via the use of diaries to record amount of Internet use and
websites visited/purchased from during the follow-up period.
5.2. Limitations and future considerations
5.3. Conclusions
There are a number of limitations with the present study which
warrant further attention. The conclusion that the training pro- Using an adapted health behaviour model, the present study ex-
gram did not affect secure behaviour can only be applied to the plored manipulating susceptibility and control as a way of promot-
generalised measure of behaviour that was adopted for the study. ing secure behaviour in Internet users. Findings indicate a positive
The findings relating to the training program may have been differ- change in both intentions to behave and actual behaviour 7 days
ent if the measure was developed to specifically focus on the later following presentation of susceptibility information which
phishing elements that the users were trained in. However, as was tailored to the user’s behaviour. The improvement was ob-
one purpose of the study was to promote secure behaviour in a real served regardless of whether the participant was told they were
world setting, participants were always likely to be exposed to a at high risk or low risk. There was no difference in reported behav-
greater, and more varied range of security issues than those faced iour between those given training to highlight a coping strategy
in the training sessions. and those not given training, raising the issue of near and far trans-
A further limitation refers to use of self-report data, as it is fer in phishing training programmes. Using the adapted health be-
noted that all data collected was reliant upon self-reports which lief model provides a wider perspective on the nature of
were requested retrospectively and are commonly prone to errors behavioural change in relation to online financial security. It ap-
(Turner & Martin, 1984). Such errors could include demand charac- pears that Internet users who complete financial transactions on-
teristics, whereby the participant responds in the way they think line can be encouraged to behave more securely in terms of how
they should, rather than as a true reflection of their actual behav- and when they disclose their personal information.
iour, i.e., participants in this study reporting that they behaved se-
curely because they know that they should. However, a number of
Acknowledgment
related studies suggest that users, despite knowing what they
should being doing, often report acting ‘inappropriately’. Over
The authors would like to thank CUPS lab at Carnegie Melon
two-thirds of ATM users reported that they would continue to
University for supporting their use of the Anti-Phishing Phil train-
use ATMS as they normally would despite their knowledge of the
ing program.
fraud risks (NCR, 2006). A qualitative study examining ATM and
Internet use found that users reported being indifferent towards
References
the threat of fraud and disregarded this information in their behav-
ioural practices. (Davinson & Sillence, in preparation). The substan- Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In J.
tial literature on the negligent use of passwords (Bishop & Klein, Kuhland & J. Beckman (Eds.), Action-control: From cognitions to behaviour
1995; Morris & Thompson, 1979) and privacy policies (Jensen, (pp. 11–39). Heidelberg: Springer.
Akbulut Sahin & Eristi (2010). Development of a scale to investigate
Potts, & Jensen, 2005; TRUSTe, 2006) also points to the fact that cybervictimization among online social utility members. Contemporary
users do not always report acting securely despite their awareness Educational Technology, 1(1), 46–59.
of appropriate practices. Akbulut, Y., Sendag, S., Birinci, G., Kilicer, K., Sahin, M. C., & Odabasi, H. F. (2008).
Exploring the types and reasons of Internet-triggered academic dishonesty
It is also worth noting that the use of self-reports is a common
among Turkish undergraduate students: Development of Internet-triggered
behaviour change paradigm that is used extensively to measure Academic Dishonesty Scale (ITADS). Computers and Education, 51(1), 463–473.
baseline, intentions and follow-up behaviour across a number of Anti-Phishing Working Group. (2007). Phishing activity trends report, December 2007.
domains, such as healthy eating (Epton & Harris, 2008), smoking <http://www.apwg.org/reports/apwg_report_dec_2007.pdf>.
Becker, M. H., Maiman, L. A., Kirscht, J. P., Haefner, D. P., & Drachman, R. H. (1977).
cessation (Hammond, Fong, McDonald, Cameron, & Brown, 2003) The health belief model and prediction of dietary compliance: A field
and exercise taken (Bock, Marcus, Pinto, & Forsyth, 2001). There- experiment. Journal of Health and Social Behavior, 18(4), 348–366.
fore, despite the limitations associated with self-report data the Becker, M. H., & Rosenstock, I. M. (1987). Comparing social learning theory and the
health belief model. In W. B. Ward (Ed.), Advances in health education and
methodology follows an established paradigm that tackles an area promotion (pp. 245–249). Greenwich, CT: JAI Press.
in which it is otherwise very difficult to obtain concrete evidence of Berthon, P. L., Pitt, L. F., & Watson, R. T. (1996). Marketing communication and the
behaviour change. world wide web. Business Horizons, 39, 43–54.
Bhattacherjee, A. (2002). Individual trust in online firms: Scale development and
It is noted that the study failed to measure Internet usage initial test. Journal of Management Information Systems, 19(1), 211–241.
behaviour in the week preceding the follow-up measure. There- Bishop, M., & Klein, D. (1995). Improving system security via proactive password
fore, it is not known what kinds of emails and websites, legitimate checking. Computers and Security, 14(3), 233–249.
Bock, B. C., Marcus, B. H., Pinto, B. M., & Forsyth, L. H. (2001). Maintenance of
or otherwise, participants were encountering. However, the demo- physical activity following an individualized motivationally tailored
graphic information collected does indicate 35% of participants intervention. Annals of Behavioral Medicine, 23, 79–87.
made five or more purchases online in the previous month, and Briggs, P., de Angeli, A., & Simpson, B. (2004). Personalisation and trust: A reciprocal
relationship? In M. C. Karat, J. Blom, & J. Karat (Eds.), Designing personalized user
56% had made between one and five purchases, therefore it is rea-
experiences for ecommerce. Kluwer.
sonable to assume they would have made an online purchase in Briones, M. G. (1998). Internet innovations—and privacy issues—remain marketing’s
the week recorded by the follow-up measure. Even if participants biggest story. Marketing News, 32(7), 1–16.
did not make any purchases doing the follow-up period it is likely Campbell, M. K., DeVellis, B. M., Strecher, V. J., Ammerman, A. S., DeVillis, R. F., &
Sandler, R. S. (1994). The impact of message tailoring on dietary behavior
that they would have browsed the Internet (95% reported using the change for disease prevention in primary care settings. American Journal of
Internet for five or more hours per week) thus providing them with Public Health, 84, 783–787.
N. Davinson, E. Sillence / Computers in Human Behavior 26 (2010) 1739–1747 1747

Chellappa, R. K., & Sin, R. (2005). Personalization versus privacy: An empirical McQuivey, J., & Ham, C. (2000). Why some young consumers don’t shop online.
examination of the online consumer’s dilemma. Information Technology and Forrester research brief, June 2, 2000.
Management, 6(2-3), 181–202. Metzger, M. J. (2004). Privacy, trust, and disclosure: Exploring barriers to electronic
Corey, S. M. (1937). Professional attitudes and actual behavior. Journal of Educational commerce. Journal of Computer-Mediated Communication, 9(4).
Psychology, 28(1), 271–280. Milletary, J. (2005). Technical trends in phishing attacks. CERT Coordination Center,
Culnan, M. J. (1999). Georgetown Internet privacy policy study: Privacy online in 1999: Carnegie Mellon University. <http://www.cert.org/archive/pdf/Phishing_trends.
A report to the Federal Trade Commission. Washington, DC: Georgetown pdf>.
University. Moore, T., & Clayton, R. (2007). An empirical analysis of the current state of phishing
Davinson, N., & Sillence, E. (in preparation). Understanding user perceptions of attack and defence. In Proceedings of the 2007 workshop on the economics of
fraud in technology mediated financial transactions. information security (WEIS2007).
Dhamija, R., Tygar, J. D., & Hearst, M. (2006). Why phishing works. In Proceedings of Morris, R., & Thompson, K. (1979). Password security: A case history.
the CHI 2005 conference of human factors in computing systems. New York: ACM Communications of the ACM, 22(11).
Press. Myers, S. (2007). Introduction to phishing. In M. Jakobsson & S. Myers (Eds.),
Downs, J. S., Holbrook, M., & Cranor, L. F. (2007). Behavioral response to phishing Phishing and countermeasures. New Jersey: John Wiley and Sons.
risk. In APWG ecrime researchers summit, Pittsburgh, PA, USA, October 4–5, Namlu, A. G., & Odabasi, F. (2007). Unethical computer using behavior scale: A study
2007. of reliability and validity on Turkish university students. Computers and
Egger, F. N. (2000). ‘‘Trust Me, I’m an Online Vendor”: Towards a model of trust for Education, 48, 205–215.
e-commerce system design. In Proceedings of the CHI 2000. ACM Press. NCR (2006). ID theft and ATM fraud, NCR Webinar Series. Available from http://
Emigh, A. (2007). Phishing attacks: Information flow and chokepoints. In M. www.hacfe.gr/events/Papers/NCR-ATM_Fraud_WP06.pdf.
Jakobsson & S. Myers (Eds.), Phishing and countermeasures. New Jersey: John Radelfinger, S. (1965). Some effects of fear-arousing communications on preventive
Wiley and Sons. health behaviour. Health Education Monographs, 19, 2–5.
Epton, T., & Harris, P. R. (2008). Self-affirmation promotes health behavior change. Riegelsberger, J., Sasse, M. A., & McCarthy, J. (2003). Shiny happy people building
Health Psychology, 27(6), 746–752. trust? Photos on e-commerce websites and consumer trust. In Proceedings of the
Festinger, L. (1957). A theory of cognitive dissonance. Evanston, IL: Row Peterson. CHI 2003. ACM Press.
Fogg, B. J., Marshall, J., Laraki, O., Osipovich, A., Varma, C., Fang, N., et al. (2001). Rippetoe, P. A., & Rogers, R. W. (1987). Effects of components of a protection
What makes web sites credible? A report on a large quantitative study. In motivation theory on adaptive and maladaptive coping with a health threat.
Proceedings of the ACM CHI 2001 conference human factors in computing systems. Journal of Personality and Social Psychology, 52, 596–604.
New York: ACM Press. Rogers, R. W. (1975). A protection motivation theory of fear appeals and attitude
Fogg, B. J., Soohoo, C., Danielson, D. R., Marable, L., Stanford, J., & Tauber, E. R. (2003). change. Journal of Psychology, 91, 93–114.
How do users evaluate the credibility of web sites? A study with over 2500 Rohm, A., & Milne, G. R. (1998). Emerging marketing and policy issues in electronic
participants. In Proceedings of the ACM CHI 2003 conference human factors in commerce: Attitudes and beliefs of Internet users. In A. Andreasen, A. Simonson,
computing systems. New York: ACM Press. & N. C. Smith (Eds.), Marketing and public policy proceedings (Vol. 8). Chicago:
Fram, E. H., & Grady, D. B. (1997). Internet shoppers: is there a surfer gender gap? American Marketing Association.
Direct Marketing, 59(9), 46–50. Rosenstock, I. M. (1966). Why people use health services. Millbank Memorial Fund
Greenwald, A. G., Carnot, C. G., Beach, R., & Young, B. (1987). Increasing voting Quarterly, 44, 94–124.
behaviour by asking people if they expect to vote. Journal of Applied Psychology, Schneier, B. (2000). Secrets and lies. New York, NY: John Wiley and Sons.
72(2), 315–318. Sheehan, K. B., & Hoy, M. G. (2000). Dimensions of privacy concern among online
Hammond, D., Fong, G. T., McDonald, P. W., Cameron, R., & Brown, K. S. (2003). consumers. Journal of Public Policy and Marketing, 19(spring), 62–73.
Impact of the graphic Canadian warning labels on adult smoking behaviour. Sheeran, P. (2002). Intention-behavior relations: A conceptual and empirical review.
Tobacco Control, 12, 391–395. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol.
IDC. (June 2008). IDC finds more of the world’s population connecting to the Internet in 12). Chichester, UK: Wiley.
new ways and embracing web 2.0 activities. Press release. <http://www.idc.com/ Sheng, S., Magnien, B., Kumaraguru, P., Acquisti, A., Cranor, L. F., Hong, J., et al.
getdoc.jsp?containerId=prUS21303808> (retrieved on 10.09.2008). (2007). Anti-Phishing Phil: The design and evaluation of a game that teaches
Internet World Stats. (August 2008). <http://www.Internetworldstats.com/ people not to fall for phish. In Proceedings of the third symposium on usable
stats.htm> (retrieved on 10.09.2008). privacy and security, Pittsburgh, PA, July 18–20.
Janis, I. (1974). Vigilance and decision making in personal crises. In G. Coelho, D. Sillence, E., Briggs, P., Harris, P., & Fishwick, L. (2006). A framework for
Hamburg, & R. J. Adams (Eds.), Coping and adaptation. New York: Basic Books. understanding trust factors in web based health advice. International Journal
Jarvenpaa, S. L., & Todd, P. A. (1996–1997). Consumer reactions to electronic of Human–Computer Studies, 64, 697–713.
shopping on the world wide web. International Journal of Electronic Commerce, Sillence, E., Briggs, P., Harris, P., & Fishwick, L. (2007). Going online for health advice:
1(winter), 59–88. Changes in usage and trust practices over the last five years. Interacting with
Jarvenpaa, S. L., & Todd, P. A. (1997). Is there a future for retailing on the Internet? In Computers, 19, 397–406.
R. A. Peterson (Ed.), Electronic marketing and the consumer. Thousand Oaks, CA: Skinner, C. S., Strecher, V. J., & Hospers, H. (1994). Physician recommendations for
Sage Publications. mammography: do tailored messages make a difference? American Journal of
Jensen, C., Potts, C., & Jensen, C. (2005). Privacy practices of Internet users: Self- Public Health, 84, 43–49.
reports versus observed behavior. International Journal of Human-Computer Spiekermann, S., Grossklags, J., & Berendt, B. (2002). Eprivacy in 2nd generation e-
Studies, 63, 203–227. commerce: Privacy preferences versus actual behavior. In Proceedings of the
Kirscht, J. P., & Haefner, D. P. (1973). Effects of repeated threatening health third ACM conference on electronic commerce – EC’01 (pp. 38–47).
communications. International Journal of Health Education, 16, 268–277. Sprott, D. E., Spangenberg, E. R., & Fisher, R. (2003). The importance of normative
Kreuter, M. W., & Strecher, V. J. (1996). Do tailored behavior change messages beliefs to the self-prophecy effect. Journal of Applied Psychology, 88(3), 423–431.
enhance the effectiveness of health risk appraisal? Results from a randomized Stanford, J., Tauber, E., Fogg, B. J., & Marable, L. (2002). Experts vs. online consumers: A
trial. Health Education Research, 11(1), 97–105. comparative credibility study of health and finance web sites. Consumer web
Kumaraguru, P., Rhee, Y., Sheng, S., Hasan, S., Acquisti, A., Cranor, L. F., et al. (2007). watch research report. <http://www.consumerwebwatch.org/news/report3_
Getting users to pay attention to anti-phishing education: Evaluation of crediilityresearch/slicedbread_abstract.htm> (retrieved on August 2003).
retention and transfer. In APWG ecrime researchers summit, Pittsburgh, PA, USA, Strecher, V. J., Kreuter, M. W., Den Boer, D. J., Kobrin, S. C., Hosper, H. J., & Skinner, C.
October 4–5, 2007. S. (1994). The effects of computer-tailored smoking cessation messages in
LaPiere, R. (1934). Attitudes versus actions. Social Forces, 13, 230–237. family practice settings. Journal of Family Practice, 39, 262–270.
LaRose, R., Rifon, N., & Enbody, R. (2008). Promoting personal responsibility for Tan, M., & Teo, T. S. H. (2000). Factors influencing the adoption of Internet banking.
Internet safety. Communications of the ACM, 51(3). Journal of Associative Information Systems, 1(5).
Leventhal, H. (1970). Findings and theory in the study of fear communications. In L. TRUSTe (2006). Consumers have a false sense of security about online privacy – Actions
Berkowitz (Ed.), Advances in experimental social psychology (Vol. 5). New York: inconsistent with attitudes. Conducted by TNS, December 2006. Available from
Academic Press. http://www.truste.org/about/press_release/12_06_06.php.
Leventhal, H. (1973). Changing attitudes and habits to reduce risk factors in chronic Turner, C., & Martin, E. (1984). Surveying subjective phenomena. New York: Russell
disease. American Journal of Cardiology, 31, 571–581. Sage Foundation.
Leventhal, H., Benjamini, Y., Brownlee, S., Diefenbach, M., Leventhal, E. A., & Patrick- Whitten, W. B., & Bjork, R. A. (1977). Learning from tests: Effects of spacing. Journal
Miller, L. (1997). Illness representations: Theoretical foundations. In K. J. Petrie of Verbal Learning and Verbal Behaviour, 16(4), 465–478.
& J. A. Weinman (Eds.), Perceptions of health and illness. Amsterdam: Harwood. Witte, K., & Allen, M. (2000). A meta-analysis of fear appeals: Implications for
Maignan, I., & Lukas, B. A. (1997). The nature and social uses of the Internet: A effective public health campaigns. Health Education and Behavior, 27(5),
qualitative investigation. Journal of Consumer Affairs, 31(2), 346–371. 591–615.
McGuire, W. (1968). The nature of attitudes and attitude change. In G. Lindzey & E. Wolhandler, H. C. (1999). Real numbers behind ‘net profits’. ActivMedia research.
Aronson (Eds.). The handbook of social psychology (Vol. 3). Reading, MA: <http://www.activmediaresearch.com/real_numbers_1999.html>.
Addison-Wesley.

You might also like