Upload
sm-furnell
View
216
Download
4
Embed Size (px)
Citation preview
i n f o rma t i o n s e c u r i t y t e c hn i c a l r e p o r t 1 5 ( 2 0 1 0 ) 4 2e4 6
ava i lab le a t www.sc iencedi rec t .com
www.compseconl ine . com/publ i ca t ions /prod in f .h tm
Online identity: Giving it all away?
S.M. Furnell
Centre for Security, Communication and Network Research, University of Plymouth, Plymouth, United Kingdom
E-mail address: [email protected]/$ e see front matter ª 2010 Elsevdoi:10.1016/j.istr.2010.09.002
a b s t r a c t
With a wealth of personal data now residing across various locations online, individuals
can find themselves at increasing risk of too much information being exposed. This in turn
may increase the potential for threats such as cyber-snooping, social engineering, and
identity theft based upon the gathered details. In many cases the exposure occurs as
a result of what individuals directly post about themselves on social networks and blog
sites, whereas in some cases it happens thanks to other people posting things beyond their
control. This paper examines the potential risks and some of the routes by which infor-
mation might be harvested. It then proceeds to consider some of the potential conse-
quences, presenting examples of how people can be duped using freely available
information and how willingly they appear to expose it to others. Recognising the ease of
online search, and the difficulty of reigning back information once it is exposed, the
requirement is clearly to improve user awareness and control over their data in the first
instance.
ª 2010 Elsevier Ltd. All rights reserved.
1. Introduction personal details. Some of the resultant implications are then
There are now significant opportunities for identities to be
reconstructed from the fragments of information that can be
foundacross various online sites and services. Careless and ill-
considered disclosure of information online can consequently
end up presenting a relatively unobstructed view for snoopers,
tricksters and thieves. Moreover, much of the details that we
find ourselves required to disclose as part of normal activities
(e.g. registration on websites) can then sit there as a potential
target for later harvesting (especially if, as in some sites,we are
required to configure associated privacy options correctly in
order to keep details from being on general view).
This paper considers the potential risks introduced by the
over-exposure and easy accessibility of personal data. The
discussion begins with an overview of selected threat cate-
gories, before proceeding to consider more specific aspects of
theproblem including the ease of locating potentially sensitive
data, the frequent vulnerability of users, and their tendency for
exacerbating this vulnerability by careless sharing of their
k.ier Ltd. All rights reserve
examined in terms of compromising other security controls
and the difficulty in regaining control of personal information
once it has been made available online.
2. Are we telling too much?
One of Berners-Lee’s original visions for the web was an envi-
ronment in which individuals could publish information as
easily as they could read it. While this did not immediately
materialise in the formative years (with farmorepeople simply
being consumers of content), more recent times have seen the
arrival of what became known as Web 2.0, with blogs, social
networks, wikis and other outlets for interactive information
sharing. With these sites and services now present in abun-
dance, many people have taken the opportunity to share their
thoughts, ideas, and indeed their lives with the rest of the
onlineworld.A good example here is providedby the growthof
social networks. The most popular, Facebook, now has in
d.
i n f o rm a t i o n s e c u r i t y t e c hn i c a l r e p o r t 1 5 ( 2 0 1 0 ) 4 2e4 6 43
excess of 500 million users (Facebook, 2010), and according to
figures from June 2010 users in both theUS and theUK are now
spending an average of over 6 h per month using the service
(representing an increase of almost 30 min per month for the
UK users and over 80 min for the US subscribers compared to
a year earlier) (BBC, 2010a). Unfortunately, however, this extra
time has not been spent learning how to protect the personal
information being disclosed. As a result, a significant risk is
that people can end up saying too much.
Here, then, is the downside. Although the opportunity
offered byWeb 2.0 and beyond is immensely empowering, the
resulting proliferation of personal and professional informa-
tion may inadvertently lend support to a variety of undesir-
able scenarios. For example, each of the following threats can
be significantly assisted by what gets posted online by (or
about) the subsequent victims:
� Snooping and cyber-stalking. Here theavailability of information
about individuals can present opportunities for unwanted
attention. For example, as a baseline there is a chance plain
and simple nosiness, with would-be snoopers being given
access to far more information than they really ought to be
entitled tosee.Atamoresinister level, thesameroutesmaybe
used to spy upon people, with details being exposed that they
might prefer to protect. As a simple experiment from the
nosiness perspective, readers could try a search that targets
old school friends, ex-workmates or current colleagues just to
see the variety and volume of information that can be
returned.
� Social engineering. In this context, the availability of infor-
mation can be leveraged towards the exploitation of a target
user. The things that people post online (or the things that
others post about them) can give numerous insights into
their background and activities, which might then be used
to trick them into accepting a subsequent request or
instruction from an attacker (be it online, by phone or in
person). In many cases, the potential to avoid being duped
rather depends upon people remembering what they have
made publically available, and not being convinced of
someone else’s legitimacy just because they can quote such
details back to them. Those who blog and tweet their every
move may still be surprised at what other people know
about them as a consequence.
� Identity theft and identity fraud. If enough information is out
there then it can be pieced together to provide a sufficient
basis for masquerading as the legitimate user. Acquiring
a home address, a date of birth and amother’smaiden name
(all of which may be available with a bit of searching), could
be sufficient to put someone in a position to take over an
account or start the process of applying for new credit in the
victim’s name. The problem has grown significantly in
recent years, and the threat has undoubtedly been amplified
by the wealth of information that is up for grabs online,
either directly or via scams such as phishing. For example,
findings from Javelin Strategy & Research suggest that over
11 million consumers fell victim to identity fraud in 2009,
compared to almost 10million in 2008 (although it is notable
that traditional methods such as physical theft still
accounted for nearly twice as many known cases as online
attacks) (Miceli and Kim, 2010).
With these problems in mind, it is worth considering how
information might be gathered, and what may be out there to
be found.
3. Digging around
Would-be data scavengers have a number of tools at their
disposal. Themostbasic is the searchengine,which canquickly
reveal a wealth of details (as well as associated images should
the latter benecessary for the intendedscam). Inaddition, there
are more specialised ‘people search’ tools (e.g. Pipl and 123
people) that specifically aim to deliver results about target
individuals. As an example, Fig. 1 shows the results of a Pipl
search for me (using details that you might easily gather, or at
least assume, from the author and affiliation details of this
paper,namely ‘StevenFurnell’ living in ‘Plymouth,GB’).Muchof
the returned information does indeed relate tome (with the list
going farbeyondthatshown inthescreenshot), albeitwithmost
of it relating to my professional activities and the only notable
personal link returned being to anold entry on the electoral roll.
Nonetheless, Pipl proved able to draw together quite relevant
summary from other sources that someone would otherwise
have needed to harvest manually.
Of course, as an academic with a variety of publications
and other external facing activities, the findings from a self-
search are not likely to be representative of what would be
found for others. With this in mind, I thought I would test the
effectiveness using the name of my best friend from school,
whom I have not seen for some 20 years. From a quick Pipl
search using just his name (a very common first name and not
a particularly uncommon surname) and the country, I was led
directly to a link showing his entire family tree, publicly
available online. This included dates of birth, dates of
marriage,maiden names, sibling namese in short, a wealth of
details that could help to inform identity theft or other forms
of masquerade attacks (especially with a variety of these
details being used in scenarios such as identity verification
and password recovery questions, as discussed later).
The family tree example highlights the fact that users do
not necessarily end up exposing themselves, but can easily find
that other people have done it for them. There are numerous
other examples that can illustrate this problem. For instance,
we can consider the various people who elect to protect their
privacy by entering a false date of birth on social networking
sites, only to find their efforts undone when a slew of close
friends then decide to post best wishes on their actual
birthday. Similarly, people who have avoided putting photos
of themselves online can find this undermined when friends
and acquaintances still tag them in photos that they have
uploaded. Meanwhile, people can also find themselves at risk
from bloggers, and from personal experience I can recount
instances ofmy travels and the birth ofmy son (completewith
photo) having been reported on other people’s blogs without
any prior checks being made with me.
In addition to opportunities to go after specific individuals,
there can also be opportunities to harvest details about
multiple users from specific sites. As an example, the news
that the aforementioned Facebook service had reached 500
million active users was soon followed by news that details of
Fig. 1 e The results of a Pipl search using the author’s name and location.
i n f o rma t i o n s e c u r i t y t e c hn i c a l r e p o r t 1 5 ( 2 0 1 0 ) 4 2e4 644
100 million of them had been collected and posted online in
a single downloadable list (compiled by a security consultant
using an automated script in order to highlight the privacy risk
of the social networking site) (Emery, 2010). Although the
collated details were all from publicly visible profile data,
commentators observed that many users expose too much
data in this context because of their difficulty with privacy
settings. Indeed, Facebook had already faced earlier contro-
versy thanks to changes it made to its privacy options (pre-
senting users with highly granular controls that many were
unable to understand or configure correctly), with the conse-
quence that many users ended up inadvertently exposing
their profile data to a wider audience than they intended (BBC,
2010b). While the published list did not directly include more
obviously personal information such as email, phone and
postal addresses, it did include the name, ID and URL of every
searchable user’s profile; enough then to provide a summary
of all the potentially exploitable profiles all in one place.
4. Vulnerable targets
As previously indicated, one of the significant risks resulting
from the abundance of available information is the recon-
naissance opportunity that it offers to social engineers. Here
the gathered details are then used in an attempt to convince or
persuade the victim of the legitimacy of the person contacting
them (i.e. cultivating the assumption that they must be
genuine because they seem so well-informed). With this in
mind, it is clear that by lettingattackers knowsomethingabout
them, individualsbecomepotentiallyvulnerable tocraftedand
targeted social engineering attempts. For example, informa-
tion taken from a web page describing their personal interests
may be exploited to promote trust based upon what Cialdini
(2000) refers to as ‘liking and similarity’ (i.e. where the
attackerexploits the fact that targetsaremore likely to respond
to someone they like, or perceive to be similar to themselves;
aspects that could clearly be mimicked if related information
about the target’s interests and the like can be found online).
Part of the problem is that people posting their details
online only tend to think about the reasons theywant to share
the information and the type of people they want to share it
with. Significant eventualities may remain unforeseen in
advance, as they simply do not recognise the potential for
misuse if the same data was to sit in different hands. It is
worth noting that this can often present a risk to the indi-
vidual and the organisation in equal measure. The latter can
most occur when someone is posting information about their
employment, which then (for example) conveys details that
might allow them to be targeted from that perspective. In
addition, they might also divulge details that the organisation
would prefer not to have shared (e.g. details of workplace
problems, or other internalmatters). So, when looking at what
someone has blogged or put on their social network page
about their work, does it look like they have applied the litmus
test of ensuring that they do not say anything more about
themselves, their role and activities within the organisation
then could be found on their employer’s own website? In
many cases the answer will be ‘no’, and this consequently
represents one area in which people can usefully be made
more aware of data value and exploitability.
Prior research has suggested that users are extremely
susceptible to exploits that are able to leverage even a small
amount of knowledge about their employer and/or workplace.
For example, a study conductedby researchers at theUniversity
of Plymouth targeted staff within a participating organisation
and used freely available information about the organisation to
fabricate a message claiming to be from their IT department.
This requested their cooperation to download a claimed soft-
ware update, and therefore served to test whether the users
concerned were vulnerable to accepting and running malware
that may well have been distributed using the same approach.
i n f o rm a t i o n s e c u r i t y t e c hn i c a l r e p o r t 1 5 ( 2 0 1 0 ) 4 2e4 6 45
Having sent the message to over 150 people, 23% were wit-
nessed to have attempted the download within the first three
and a half hours (Furnell and Papadaki, 2008.). The fact that the
message to staff was deliberately worded and presented in
a manner that ought to have raised suspicion (and yet snared
a much greater number than apparently reported it to the IT
department) lends further weight to the assertion that many
users will require very little convincing in order to be converted
from potential to actual victims.
5. Sharing without caring
Some people not only put too much information online, they
also do things that increase their own risk of exposing it. For
example, many users on social networking sites will admit to
having accepted friend requests from relative (or total)
strangers; andhaving shared their full profile datawith them in
theprocess. The practice is perhapsmost pronouncedamongst
youngpeople,where theobjective isoftennot tomaintainor re-
establish contactwith friends, but rather to collectmore online
acquaintances than their peers. Thus, issuing or accepting
friend requestswith strangers is innowayunusual andhelps to
enhance their apparent popularity rating. Indeed, a poll con-
ducted amongst a group of 16e17-year-old pupils at an infor-
mation security workshop revealed that 59% had invited
someone they had nevermet to become their friend on a social
network, and 47%had accepted a similar request received from
anonline stranger (Furnell, 2008). In addition, 28%had used the
site to find things about other people (e.g. personal information
or details of their activities), illustrating that if the opportunity
for snooping is there thenmany people will take it.
In order to get a more practical measure of people’s
tendency to give their data away, a further study from the
University of Plymouth set out to determine the ease with
which such data could be gathered (Phippen et al., 2009). The
approach involved issuing friend requests to users on a social
networking site in order to determine howmanywould accept
such a request from a stranger and thereby share their own
personal data. In order to provide a basis for inviting others,
fake social network profiles were created for two characters,
one male and one female, including a range of false personal
information.While details such as phone number and address
were checked to ensure that they were not real, the name of
a real school was used as this enabled the social networking
site to automatically suggest other potential friends who had
attended the same place. With the fake profiles set to private
(i.e. so that only people who were friends with them could see
any significant details beyond the name and photo), invita-
tions were then sent to 100 people per profile (50 to males and
50 to females). In order to improve the chances of soliciting
interest, provocative pictures were assigned to both profiles
(specifically, one depicting a bare chest and six-pack for the
male and a large cleavage for the female). Neither photo
included any headshot, and so all that the inviteeswould have
seen would have been invitations from someone seeming to
have attended the same school as them.
After amonth, the 57 of the requests issued on behalf of the
female character had been accepted (48 of these acceptances
occurring within the first 24 h), compared with 15 for the male
character. Notably, there were no reminders following the
original invitation, andboth achieveda fairlywell balancedmix
of friends of each gender. As a consequence of the accepted
invitations, the new friends were all sharing potentially harm-
ful personal details with complete strangers. Themajority was
divulging birthday, relationship details, photos, activities and
interests, and email contacts, with small proportions also
sharing physical addresses, telephone numbers and employ-
ment details. It is also worth noting that in addition to the
accepted invitations, the female character also received 29
friend requests of her own, suggesting the apparent ease with
which users can be taken in by a good social engineering hook.
6. Answers to personal questions
Looking at the above results, some users would still see little
problem, in the sense direct security details such as account
codes and passwords are not being disclosed. Indeed, it is
sometimes difficult to appreciate why sharing a few tidbits of
personal information canbeall that problematic. Thedifficulty
is, of course, is the different contexts in which such informa-
tion now gets used. For example, and as mentioned earlier,
personal details are often employed as the basis for questions
used in website password recovery and reset procedures. The
risk here ultimately comes down to a combination of howwell
the security questionshave been chosenandhowmuch canbe
found out about the user in order to answer them. In many
cases such challenges ask the user for one or more items of
personal information, and so there is clearly a risk if the related
answers can be researched from sources such as social
networks, blogs, and genealogy services.
Choosing the questions that should form the basis of the
challenges has always merited special consideration, in order
to avoid basing them around information that would be
known or easily discoverable by someone else, as well as to
ensure that they require answers that the user would be likely
to remember. As such, factual questions are typically prefer-
able to those based around factors such as the user’s prefer-
ences or word associations (both of which may change over
time, with the consequent risk that the user may forget how
they originally answered when registering their responses)
(Haga and Zviran, 1991). However, the factual responses are
the ones that might be more easily researched from the user’s
personal web pages etc., especially if they are prone to
disclosing too much about themselves online.
With this in mind, we can compare the two sets of ques-
tions in Fig. 2. The first set comes from eBay, and it is clear that
a number of themmight be researchable from online sources.
By contrast, the other set, taken from Windows Live, would
appear to be less susceptible to compromise in this context.
The user’s supposedly secret responses can be especially
vulnerable if questions are framed around current information
or interests, as these are more likely to be the types of details
that users are prone to divulging in their social networks and
blog pages. So, from the perspective of potential compromise,
there is a tangible difference between asking questions such
as ‘What is your pet’s name?’ and ‘Name of first pet’, with the
former having a significantly greater likelihood of being
discovered through users’ social networking postings than the
Fig. 2 e Secret question options from (a) eBay and (b)
Windows Live.
i n f o rma t i o n s e c u r i t y t e c hn i c a l r e p o r t 1 5 ( 2 0 1 0 ) 4 2e4 646
latter (unless the user concerned is posting extensively about
their life history etc.).
7. Conclusions
If sensitive data has found its way onto the web then the
chances are that it is going to stay there, even if the user that
originally posted it changes their mind about making it avail-
able. Efforts to retrospectively restoreprivacy canbe frustrated
by the persistence and sheer accessibility of online informa-
tion; indeed the very factors that are so advantageous in other
search contexts only serve to amplify the risk if something has
gone awry. So, while information can often be removed from
a site, it is far less likely to be removed from the web, and with
search engines trawling cyberspace as a whole there is a good
chance of it being relocated.
With the above in mind, the solution is to try to control
what gets out there in the first place. Control in this context
should not mean restricting or filtering what users can do,
but rather making them more aware of the onus upon them
to police and regulate their own activities. Of course, the
experiences to date clearly illustrate that many will not be in
a position to do this without support. There is certainly
a need to boost user awareness of data value when consid-
ering what and where they are willing to share. In addition,
the technology needs to make it as easy for them to protect
information as it does to publish it. This means privacy
controls that are clearly and (where possible) consistently
presented, so that users have a chance to learn good practice
and then put it to use across the range of sites they use. Thus,
as with many other aspects of security, it is the combination
of technology and awareness that holds the key to address-
ing the problem.
r e f e r e n c e s
BBC. The ups and downs of social networks. BBC News Online;2010a. 22 July 2010.
BBC. Facebook privacy settings to be made simpler. BBC NewsOnline; 2010b. 26 May 2010.
Cialdini RB. Influence: science and practice. 4th ed. Allyn & Bacon;2000.
Emery D. Details of 100 m Facebook users collected and publishedby Daniel Emery. BBC News Online; 2010. 29 July 2010.
Facebook. Statistics e Press room, http://www.facebook.com/press/info.php?statistics; 2010 (accessed 27.07.10).
Furnell S. End-user security culture: a lesson that will never belearnt? Computer Fraud & Security; 2008:6e9. April 2008.
Furnell S, Papadaki M. Testing our defences or defending ourtests: the obstacles to performing security assessment.Computer Fraud & Security; 2008:pp8e12. May 2008.
Haga WJ, Zviran M. Question-and-answer passwords: anempirical evaluation. Information Systems 1991;16(3):335e43.
Miceli D, Kim R. 2010 identity fraud survey report: consumerversion. Javelin Strategy & Research; 2010. February 2010.
Phippen A, Davey R, Furnell S. Should we do it just because wecan? Methodological and ethical implications for informationrevelation in online social networks. MethodologicalInnovations Online 2009;4(3):41e55.