Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
Research paper
To share or not to share: a behavioral
perspective on human participation in security
information sharing
Alain Mermoud
1,2,
*, Marcus Matthias Keupp
2,3
,Ke´ vin Huguenin
1
,
Maximilian Palmie´
4
and Dimitri Percia David
1,2
1
Department of Information Systems, Faculty of Business and Economics (HEC Lausanne), University of Lausanne
(UNIL), 1015 Lausanne, Switzerland;
2
Department of Defence Management, Military Academy (MILAC) at ETH
Zurich, 8903 Birmensdorf, Switzerland;
3
University of St. Gallen, Dufourstrasse 50, 9000 St. Gallen, Switzerland;
4
Institute of Technology Management, University of St. Gallen, Dufourstrasse 40a, 9000 St. Gallen, Switzerland
*Corresponding address: Tel: +41 58 484 82 99; E-mail: alain.mermoud@gmail.com
Received 21 July 2018; revised 17 May 2019; accepted 14 June 2019
Abstract
Security information sharing (SIS) is an activity whereby individuals exchange information that is
relevant to analyze or prevent cybersecurity incidents. However, despite technological advances
and increased regulatory pressure, individuals still seem reluctant to share security information.
Few contributions have addressed this conundrum to date. Adopting an interdisciplinary
approach, our study proposes a behavioral framework that theorizes how and why human behav-
ior and SIS may be associated. We use psychometric methods to test these associations, analyzing
a unique sample of human Information Sharing and Analysis Center members who share real se-
curity information. We also provide a dual empirical operationalization of SIS by introducing the
measures of SIS frequency and intensity. We find significant associations between human
behavior and SIS. Thus, the study contributes to clarifying why SIS, while beneficial, is underutil-
ized by pointing to the pivotal role of human behavior for economic outcomes. It therefore extends
the growing field of the economics of information security. By the same token, it informs managers
and regulators about the significance of human behavior as they propagate goal alignment and
shape institutions. Finally, the study defines a broad agenda for future research on SIS.
Key words: security information sharing; psychometrics; economics of information security; behavioral economics, behavioral
psychology
Introduction
Security information sharing (SIS) is an activity whereby individuals ex-
change information that is relevant to analyze or prevent cybersecurity
incidents. Such information includes, but is not limited to, the identifi-
cation of information system vulnerabilities, phishing attempts, mal-
ware, and data breaches, as well as results of intelligence analysis, best
practices, early warnings, expert advice, and general insights [67].
Prior research has proposed that SIS makes every unit of security
investmentmore effective, such that individuals can reduce invest-
ments dedicated to generate cybersecurity in their organization. As a
result of these individual improvements, total welfare is also likely
to increase [41,47]. Hence, SIS likely contributes to strengthening
the cybersecurity of firms, critical infrastructures, government, and
society [19,45,46,48,54].
However, these theoretical expectations hardly seem to material-
ize. Recent contributions have noted that SIS is at suboptimal levels,
implying negative consequences for the cybersecurity of organizations
and society [19]. Game-theoretic simulation suggests that individuals
may free-ride on the information provided by others while not sharing
any information themselves [47,55]. Researchers and international
V
CThe Author(s) 2019. Published by Oxford University Press. 1
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs licence (http://creativecommons.org/ licenses/
by-nc-nd/4.0/), which permits non-commercial reproduction and distribution of the work, in any medium, provided the original work is not altered or transformed in any way,
and that the work is properly cited. For commercial re-use, please contactjournals.permissions@oup.com
Journal of Cybersecurity, 2019, 1–13
doi: 10.1093/cybsec/tyz006
Research paper
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
organizations have been warning for years that individuals seem reluc-
tant to share security information, although the technical infrastruc-
ture for information exchange does exist [32,33,47,73]. Legislators
have attempted to resolve this problem by creating regulation that
makes SIS mandatory.
1
However, reviews suggest that despite these
attempts, individuals still seem reluctant to share security information
[16,44,72,103]. They may even ‘game’ the system in an attempt to
circumvent regulation [5,71,72].
All these findings imply that human behavior may be significantly
associated with the extent to which SIS occurs (if at all). It is therefore
not surprising to see recent work emphasizing that the study of human
behavior is key to the understanding of SIS [19]. More specifically, this
work predicts that SIS can only be imperfectly understood unless the
human motivation to (not) participate in SIS is studied [53,65,98].
However, few contributions have addressed this research gap to
date. Since an excellent account of the SIS literature exists [64], we re-
frain from replicating this account here. We rather point to the fact that
this account shows that very few empirical studies on non-public SIS
exist. These few studies concentrate on analyzing incident counts and ag-
gregate data, but they do not study human behavior at the individual
level of analysis (see Ref. [64] for a tabulated overview).
Our study intends to address this gap by proposing how and why
human behavior and SIS may be associated, and by providing an em-
pirical test of this association. Following prior recommendations [6],
we adopt an interdisciplinary approach. Recently, interdisciplinary
studies were productive in showing the extent to which human behav-
ior is associated with knowledge sharing [87,106].
We build a theoretical framework anchored in behavioral theory,
arguing that SIS is associated with human behavior. We use psychomet-
ric methods to test these associations, analyzing a unique sample of 262
members of an Information Sharing and Analysis Center (ISAC) who
share real security information. The remainder of this article is struc-
tured as follows. Section 2 develops the behavioral framework and
deducts testable hypotheses from this framework. Section 3 details the
sampling context, measures, and empirical methods. The results are
explained in Section 4. Section 5 discusses both the theoretical, empiric-
al, and practical contributions our study makes and points to some limi-
tations of our approach that open up paths for future research.
Theoretical Framework and Hypotheses
Behavioral research relativized some of the strong formal assump-
tions that neoclassical economics had ascribed to human behavior,
particularly those of rationality, perfect information, and selfish util-
ity maximization (“homo oeconomicus”). In contrast, it showed
that human beings have bounded instead of perfect rationality. They
often violate social expectations, have limited information-process-
ing capacity, use heuristics when making decisions, are affected by
emotion while doing so, and retaliate even if the cost of retaliation
exceeds its benefits [13,27,37,58,59,89].
Moreover, humans do not necessarily maximize higher level (i.e.
organizational, societal) goals, even if it would be economically ra-
tional for them to do so. Theoretical work on SIS has suggested early
that individual and organizational interests may not always be
aligned and that the individual is not necessarily an indifferent agent
[42]. Goal-framing theory suggests that individual goals may not ne-
cessarily be congruent with higher level goal frames, implying that
the individual can defect from organizational maximization goals
[66]. Particularly in the case of collective action, the individual may
behave in ways that are not conducive to the overall group goal
[78,79]. For the context of SIS, this research implies that individual-
ly, humans might not necessarily participate in SIS although it would
be optimal to do so for society as a whole.
Particularly, human exchange relationships are not necessarily char-
acterized by rational economic optimization, but instead by human
expectations about fairness, reciprocity, and trust [36,37,39,68].
Therefore, the argument can be made that SIS may be associated with
human behavior. Indeed, prior research argues that the understanding
of SIS requires an analysis of what behavior may motivate humans to
participate in SIS and what may deter them from doing so [8,10].
Human behavior is the result of human motivation, intention,
and volition. It manifests itself in goal-directed (i.e. nonrandom) and
observable actions [90,93,102]. Sharing information implies
human action from at least the side of the individual who shares.
Moreover, SIS constitutes an economic transaction by which know-
ledge resources are shared, rather than acquired [17]. Hence, SIS dif-
fers from discrete arm’s length transactions, whereby a single
individual simply trades financial means for access to information.
Instead, SIS is characterized by continued social interaction among
many individuals who mutually exchange information assets [106].
Therefore, humans are unlikely to randomly participate in SIS,
such that SIS does not occur “naturally.” Hence, theorizing is
required regarding how and why human behavior may be associated
with SIS. Applying prior behavioral research to our research con-
text, we develop testable hypotheses about five salient constructs
which may be associated with SIS. In all of these hypotheses, our
focal individual is an indifferent individual who, independently of
the motives of other individuals, ponders whether or not to partici-
pate in SIS. We believe this perspective is conservative and condu-
cive to empirical analysis since it neither requires assumptions about
the behavior of other individuals nor a dyadic research setting.
Attitude
Behavioral theory suggests that attitudes have a directive influence
on human behavior [1]. Attitude is a psychological tendency that is
expressed by evaluating a particular entity with some degree of favor
or disfavor [30]. Hence, an individual’s favorable or unfavorable at-
titude towards a particular behavior predicts the extent to which
this behavior actually occurs [2,3].
Much empirical work has confirmed and detailed this attitude–
behavior link, particularly in the context of information systems
adoption and intention to use (see Refs [14]and[62] for extensive lit-
erature reviews). More specifically, this attitude-behavior link influ-
ences individuals’ intention to share knowledge [17]. Moreover, an
affirmative attitude towards knowledge sharing positively influences
participation rates [87]. Descriptive work has conjectured (though
not tested or confirmed) that individual attitudes about the meaning-
fulness of SIS might be associated with actual participation in SIS
[32]. Therefore, if the focal individual has a positive attitude towards
SIS, s/he should be more likely to participate in SIS. Therefore,
H1: SIS is positively associated with the extent to which the focal
individual has a positive attitude towards SIS.
1 For example, the USA created the 2002 Sarbanes-Oxley Act and the
2015 Cybersecurity Information Sharing Act (CISA). The Health
Insurance Portability and Accountability Act (HIPAA) requires organiza-
tions to report breaches of protected health information (PHI) to the U.S.
Department of Health and Human Services (HHS). In December 2015,
the European Parliament and Council agreed on the first EU-wide legisla-
tion on cybersecurity by proposing the EU Network and Information
Security (NIS) Directive.
2Journal of Cybersecurity, 2019, Vol. 5, No. 1
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
Reciprocity
Behavioral theory suggests that human behavior is characterized by
inequity aversion [39]. As they socially interact with others, humans
expect to receive equitable compensation whenever they voluntarily
give something to others, and they punish those unwilling to give
something in return [22,95]. Hence, when humans are treated in a
particular way, they reciprocate, that is, they respond likewise [36].
As a result, reciprocity is a shared behavioral norm among human
beings that governs their social cooperation [38,50].
Economic exchange relationships are therefore shaped by the
reciprocity expectations of the participants involved in this exchange
[61]. In such relationships, reciprocity is a dominant strategy that is
conducive to a socially efficient distribution of resources [7,20].
Therefore, the extent to which the focal individual participates in in-
formation exchange is likely associated with that individual’s ex-
pectation that his/her efforts are reciprocated.
For example, reciprocal fairness is an important variable in the
design of peer selection algorithms in peer-to-peer networks. By inte-
grating reciprocal response patterns such as “tit-for-tat,” operators
can optimize peer-to-peer traffic [101]. The value of a unit of secur-
ity information is proportional to the incremental security enhance-
ment that this unit is supposed to provide to the recipient [18,49].
Hence, whenever the focal individual shares such information units,
it creates value for the counterparty. By the above arguments, the
focal individual likely refuses to participate in future exchanges un-
less such value creation is reciprocated by the counterparty.
On the one hand, the focal individual may expect that informa-
tion sharing is reciprocated by “hard rewards,” that is, in monetary
terms, by a higher status inside the ISAC or his or her own organiza-
tion, or in terms of career prospects (transactional reciprocity). On
the other hand, the focal individual may also expect that whenever
s/he shares a unit of information, s/he receives useful information in
return, such that a continuous social interaction that is beneficial to
both parties emerges (social reciprocity). Prior research suggests that
both these types of reciprocity are associated with information ex-
change patterns between individuals [63,80,88]. Therefore,
H2a: SIS is positively associated with the extent to which the
focal individual expects his or her information sharing to be
transactionally reciprocated.
H2b: SIS is positively associated with the extent to which the
focal individual expects his or her information sharing to be so-
cially reciprocated.
Executional Cost
Behavioral theory suggests that humans are loss-averse, that is, they
attempt to avoid economic losses more than they attempt to realize
economic benefits. Much experimental research has confirmed this
tendency [58,59,92,96,97].
An economic exchange relationship can be fraught with signifi-
cant transaction cost, i.e. the time, material, and financial resources
that the focal individual must commit before an exchange is made
[104]. Hence, if SIS is associated with high transaction costs for par-
ticipation, the focal individual is likely to avoid the necessary re-
source commitments to finance this cost. For example, Ref. [106]
argue that when knowledge contribution requires significant time,
sharing tends to be inhibited. Consistent with their conceptualiza-
tion, we term such transaction costs “executional cost.”
As a result, in the presence of high executional cost, the focal in-
dividual likely adapts his or her behavior in an attempt to avoid
these costs. For instance, if the focal individual learns that in a given
ISAC environment, SIS is taking too much time, is too laborious, or
requires too much effort, the individual likely reduces or terminates
participation in SIS [67]. For example, an abundance of procedural
rules that govern the processing and labelling of shared information
and the secure storage and access to shared data likely stalls infor-
mation sharing activity [33]. Thus, high executional cost likely dis-
suades the focal individual from participating in SIS. Therefore,
H3: SIS is negatively associated with the extent to which the focal
individual expects information sharing to be fraught with execu-
tional cost.
Reputation
Behavioral theory suggests that humans deeply care about being rec-
ognized and accepted by others [11,15]. Many philosophers have
argued that the desire for social esteem fundamentally influences
human behavior and, as a result, economic action [21].
Depending on the outcomes of particular social interactions with
other individuals, the focal individual earns or loses social esteem.
Hence, over time each individual builds a reputation, that is, a socially
transmitted assessment by which other individuals judge the focal indi-
vidual’s social esteem [31,69]. For example, academic researchers strive
to increase the reputation of their department by publishing scholarly
work [60].Thedesiretoearnareputationasacompetentdeveloperisa
strong motivator for individuals to participate in open source software
development although they receive no monetary compensation for the
working hours they dedicate to this development [99].
When this reasoning is transferred to the context of SIS, the focal
individual may be inclined to share information because s/he hopes
to build or improve his or her reputation among the other partici-
pants of SIS. Prior research suggests that this desire constitutes an
extrinsic motivation that may be associated with an individual’s in-
tention to share information [25,81], and intention is a precursor of
behavior. Therefore,
H4: SIS is positively associated with the extent to which the focal
individual expects information sharing to promote his or her
reputation in the sharing community.
Trust
Behavioral theory suggests that humans simplify complex decision-
making by applying heuristics [82,97], particularly when they attempt
to reduce the cost of information acquisition and valuation [40].
Whenever a focal individual is unable or unwilling to object-
ively evaluate information conveyed by other individuals, s/he
likely resorts to heuristics to simplify the evaluation process [24].
In the context of SIS, this implies that whenever the focal individ-
ual receives security information from another individual, s/he
cannot necessarily be sure about the extent to which (if any) this
information is valuable or useful. This assessment is associated
with significant transaction cost, for example, for due diligence
procedures that attempt to value the information received. The in-
dividual may also lack technological competence and expertise,
such that time-consuming discussions with experts are required
for proper valuation. All in all, upon the receipt of a particular
unit of information, the focal individual is faced with a complex
valuation problem which s/he may seek to simplify by applying
heuristics.
Trust is an implicit set of beliefs that the other party will behave
in a reliable manner [43]. This set of beliefs is a particularly effective
heuristic because it can reduce the transaction cost associated with
this valuation. If the focal individual trusts the information received
Journal of Cybersecurity, 2019, Vol. 5, No. 1 3
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
is useful and valuable, s/he can simplify evaluation procedures, and
particularly so if the involved individuals interact in dense networks
with agreed standards of behavior. Therefore, trust is a facilitator of
economic organization and interaction [51,70]. For example, mu-
tual trust among the participants of peer-to-peer networks can re-
duce transactional uncertainty [105]. Moreover, trust can mitigate
information asymmetry by reducing transaction-specific risks [9]. It
is also a significant predictor of participation in virtual knowledge
sharing communities [86].
Such trust, in turn, is positively associated with knowledge shar-
ing in both direct and indirect ways [56], whereas distrust is an obs-
tacle to knowledge sharing [4]. More specifically, trust is a
facilitator in information security knowledge sharing behavior [87].
Thus, the extent to which the focal individual trusts the information
s/he receives is valuable should be positively associated with his or
her propensity to participate in SIS. Therefore,
H5: SIS is positively associated with the extent to which the focal
individual trusts that the counterparty provides valuable
information.
Interaction Effects
By consequence, we suggest that trust negatively moderates the asso-
ciations between attitude and reciprocity on the one hand and SIS
on the other hand. We argued that trust is a facilitator of economic
exchange. In other words, trust likely reduces the focal individual’s
perceived cost of engaging in SIS, in that s/he requires fewer or lesser
alternative stimuli [66]. A neutral focal individual who has not par-
ticipated in SIS before is unlikely to participate unless s/he has a
positive attitude towards SIS. That individual must hence construct
the meaningfulness of SIS “internally,” that is, convince him- or her-
self that SIS is useful. By contrast, if the focal individual trusts that
the information s/he receives will be useful, s/he uses the counter-
party to “externally” confirm such meaningfulness of SIS. The pro-
cess of the internal construction of the meaningfulness of SIS is
therefore at least partially substituted by the external, trust-based af-
firmation of such meaningfulness. We would hence expect that the
significance of the association between attitude and SIS decreases
with the extent to which the focal individual trusts the information
s/he receives will be useful.
By the same token, since trust is a facilitator of economic ex-
change, it likely reduces the association between reciprocity and SIS.
An indifferent focal individual cannot be completely sure about the
behavior of the exchange counterparty, such that s/he requires con-
tinuous transactional or social reciprocity for SIS to perpetuate the
exchange. In the absence of any trust that the information received
is useful, SIS likely ends as soon as this reciprocity requirement is no
longer met. In contrast, whenever the focal individual trusts that the
information s/he receives will be useful, s/he has a motive to partici-
pate in SIS that is independent of such reciprocity concerns. Hence,
trust is likely to act at least partially as a substitute for reciprocity,
such that the focal individual should emphasize to a lesser extent
that reciprocity will be required if s/he is expected to begin or per-
petuate SIS. Therefore,
H6a–c: The extent to which the focal individual trusts that infor-
mation received from the counterparty is effective negatively
moderates the respective positive associations between attitude,
transactional, and social reciprocity on the one hand and SIS on
the other hand.
Methods
Sampling Context and Population
Our study focused on the 424 members of the closed user group of
the Swiss national ISAC, the “Reporting and Analysis Centre for
Information Assurance” (MELANI-net). An ISAC is an organization
that brings together cybersecurity managers in person to facilitate
SIS between operators of critical infrastructures. For a general intro-
duction to the concept of an ISAC, see Ref. [107]. For some illustra-
tive examples of ISACs across different countries, see Ref. [34]. For
a detailed description of MELANI-net, its organization, and history,
see Ref. [29]. The ISAC we study is organized as a public-private
partnership between the government and private industry; it oper-
ates on a not-for-profit basis. Membership in MELANI-net is volun-
tary. In Switzerland, there is no regulation that makes SIS
mandatory; hence, individuals are free to share or not share infor-
mation, and they can also control the group of individuals with
whom they want to share the information. This implies our study
design can capture the full range of human behavior from perfect co-
operation to total refusal.
The members of the closed user group are all senior managers in
charge of providing cybersecurity for their respective organizations.
They come from both private critical infrastructure operators and
from the public sector. They have to undergo government identifica-
tion and clearance procedures as well as background checks before
being admitted for ISAC membership. They share classified, highly
sensitive information the leaking or abuse of which may cause sig-
nificant economic damage. There is no interaction of these members
with the public whatsoever, and no external communication to the
public or any publication of SIS results is made. For all of these
members, the exchange of SIS can be assumed to be relevant, as they
manage critical infrastructures that are ultimately all connected and
operate with similar IT systems, such that cybersecurity problems
that relate to any particular individual are likely of interest to other
participants too.
Within this closed user group, individuals can contact each other
by an internal message board whenever a particular individual has
shared information about a threat that is of interest to other mem-
bers. They do so by commenting on the initial information shared in
order to establish a first contact, which then leads to further social
exchange between the two individuals. Once contact is made by a
short reply to the threat information, the individuals involved in the
conversation meet on their own initiative to share detailed security
information between them (e.g. informally over lunch, in group
meetings, or small industry-specific conferences, but always face-to-
face). Each individual decides for him- or herself if s/he wants to
meet, with whom, and in what form. They also freely decide about
the extent of the information shared (if any). MELANI-net officials
neither force nor encourage individuals to interact; both in terms of
social interaction in general and regarding the sharing of any par-
ticular unit of information.
Measures
Our study analyzes human behavior on the individual level of ana-
lysis. We therefore chose a psychometric approach to operationalize
our constructs [77]. We adopted psychometric scales from the ex-
tant literature wherever possible and kept specific adaptions to our
population context to a minimum. Table 1explains and details all
variables, their item composition and wording (if applicable),
dropped items (if any), factor loadings, and Cronbach alphas and
cites the sources they were taken from.
4Journal of Cybersecurity, 2019, Vol. 5, No. 1
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
SIS is operationalized dually by the two constructs “frequency”
and “intensity.” Intensity measures the extent to which the focal in-
dividual reacts to any threat information shared by another individ-
ual and thus begins social interaction with that other individual.
Intensity is thus a reactive measure of how intensely the focal indi-
vidual engages in knowledge sharing with others upon being
informed of a threat.
2
Since information sharing is not mandatory,
this measure captures the individual’s free choice to (not) engage in
exchange relationships with other individuals. In contrast, frequency
is a proactive measure; it captures how often an individual shares se-
curity information that s/he possesses him- or herself.
To capture respondent heterogeneity, we controlled for gender,
age, and education level. Further, we controlled for the individual’s
ISAC membership duration in years, because a respondent’s sharing
activity may co-evolve with the length of ISAC membership.
“Gender” was coded dichotomously (male, female). “Age” was cap-
tured by four mutually exclusive categories (21–30, 31–40, 41–50,
50þyears). “Education” was captured by six mutually exclusive
categories (none, bachelor, diploma, master, PhD, other). We also
controlled for the industry affiliation of the organization that the in-
dividual represents and combined these into five categories (govern-
ment, banking and finance, energy, health, telecom and IT, all
others).
Implementation
Data for all variables were collected from individual respondents by
a questionnaire instrument. We followed the procedures and recom-
mendations of Ref. [28] for questionnaire design, pre-test, and im-
plementation. Likert-scaled items were anchored at “strongly
disagree” (1) and “strongly agree” (5) with “neutral” as the mid-
point. Categories for the measure “intensity” were ordered
hierarchically.
The questionnaire was developed as a paper instrument first. It
was pre-tested with seven different focus groups from academia and
the cybersecurity industry.
3
Feedback obtained was used to improve
the visual presentation of the questionnaire and to add additional
explanations. This feedback also indicated that respondents could
make valid and reliable assessments.
Within the closed user group, both MELANI-net officials and
members communicate with each other in English. Switzerland has
four official languages, none of which is English, and all constructs
we used for measurement were originally published in English. We
therefore chose to implement the questionnaire in English to rule
out any back-translation problems. Before implementation, we con-
ducted pre-tests to make sure respondents had the necessary lan-
guage skills. The cover page of the survey informed respondents
about the research project and our goals and also made clear that
we had no financial or business-related interest.
The paper instrument was then implemented as a web-based sur-
vey using “SelectSurvey” software provided by the Swiss Federal
Institute of Technology Zurich. For reasons of data security, the sur-
vey was hosted on the proprietary servers of this university. The
management of MELANI-net invited all closed user group members
to respond to the survey by sending an anonymized access link, such
that the anonymity of respondents was guaranteed at all times.
Respondents could freely choose whether or not to reply. As a re-
ward for participation, respondents were offered a research report
free of charge that summarized the responses. Respondents could
freely choose to save intermediate questionnaire completions and re-
turn to the survey and complete it at a later point in time.
The online questionnaire and the reminders were sent to the
population by the Deputy Head of MELANI-net together with a let-
ter of endorsement. The survey link was sent in an e-mail describing
the authors, the data, contact details for IT support, the offer of a
free report, and the scope of our study. Data collection began on 12
October 2017 and ended on 1 December 2017. Two reminders were
sent on 26 October and 9 November 2017. Of all 424 members,
262 had responded when the survey was closed for a total response
rate of 62%.
Analysis
Upon completion of the survey, sample data were exported from the
survey server, manually inspected for consistency and then con-
verted into a STATA dataset (Vol. 15) on which all further statistical
analysis was performed. Post-hoc tests suggested no significant influ-
ence of response time on any measure. There was no significant
overrepresentation of individuals affiliated with any particular or-
ganization, suggesting no need for a nested analytical design.
We performed principal component factor analysis with oblique
rotation on all items. Validity was tested by calculating item-test,
item-rest, and average inter-item correlations. Reliability was meas-
ured by Cronbach alpha. High direct factor-loadings and low cross-
loadings indicate a high degree of convergent validity [52]. The final
matrix suggested seven factors with an eigenvalue above unity. The
first factor explained 14.56% of the total variance, suggesting the
absence of significant common method variance in the sample [84].
The detailed factor-loadings and their diagnostic measures are given
in Table 2. Upon this analysis, three items were dropped (viz.
Table 1) because they had low direct and high cross factor loadings.
Finally, for any scale, individual item scores were added, and this
sum was divided by the number of items in the scale [85,94].
The construct intensity is ordered and categorical, therefore we
estimated ordered probit models. A comparison with an alternative
ordered logit estimation confirmed the original estimations and indi-
cated the ordered probit model fit the data slightly better. The con-
struct frequency is conditioned on values between 1 and 5, therefore
we estimated Tobit models. Both models were estimated with robust
standard errors to neutralize any potential heteroscedasticity.
Consistent with the recommendation of Ref. [26], we incrementally
built all models by entering only the controls in a baseline model
first, then added the main effects, and finally entered the interaction
effects. In both estimations, we mean centered the measures before
entering them into the analysis. Model fit was assessed by repeated
comparisons of Akaike and Bayesian information criteria between
different specifications. Since all the categorical controls age, educa-
tion and industry are exhaustive and hence perfectly collinear, Stata
automatically chose a benchmark category for each of these (cf.
footnotes b to Tables 5 and 6).
2 The measure intensity is ordered and categorical in that it asks respond-
ents to provide an estimate rather than an exact percentage figure. We
preferred this approach in order to give respondents an opportunity to
provide an estimate, such that they would not be deterred by the need to
provide an exact figure. We also captured an alternative measure of in-
tensity by a Likert scale, but found that models with the ordered
categorical measure fit the data better. We also contrasted the Tobit
model that used the scale-based measure for frequency with an alterna-
tive ordered probit model that used a categorical specification of that
variable, but found that the former model fit the data much better.
3 Further detailed information about these pre-tests is available from the
corresponding author.
Journal of Cybersecurity, 2019, Vol. 5, No. 1 5
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
Table 1: Constructs, items, and scales used in the survey
Measures
(source)
Type Item Text Factor
loading
Cronbach
alpha
SIS constructs
Intensity of SIS (novel) Ordered
categorical
measure
n/a How often do you comment on shared information?
•Never
•Rarely, in less than 10% of the chances when I could have
•Occasionally, in about 30% of the chances when I could have
•Sometimes, in about 50% of the chances when I could have
•Frequently, in about 70% of the chances when I could have
•Usually, in about 90% of the chances I could have
•Every time
n/a n/a
Frequency of SIS [87] Likert scale ISKS1 I frequently share my experience about information security with
MELANI
0.8075 0.8945
ISKS2 I frequently share my information security knowledge with MELANI 0.8903
ISKS3 I frequently share my information security documents with MELANI 0.8850
ISKS4 I frequently share my expertise from my information security training
with MELANI
0.8600
ISKS5 I frequently talk with others about information security incidents and
their solutions in MELANI workshops
0.6898
Behavioral constructs
Attitude [87] Likert scale AT1 I think SIS behavior is a valuable asset in the organization Dropped 0.6761
AT2 I believe SIS is a useful behavioral tool to safeguard the organization’s
information assets
0.7751
AT3 My SIS has a positive effect on mitigating the risk of information se-
curity breaches
0.6376
AT4 SIS is a wise behavior that decreases the risk of information security
incidents
0.7849
Transactional reciprocity [100] Likert scale HR1 I expect to be rewarded with a higher salary in return for sharing
knowledge with other participants
0.8822 0.7956
HR2 I expect to receive monetary rewards (i.e. additional bonus) in return
for sharing knowledge with other participants
0.8743
HR3 I expect to receive opportunities to learn from others in return for
sharing knowledge with other participants
Dropped
HR4 I expect to be rewarded with an increased job security in return for
sharing knowledge with other participants
0.7499
Social reciprocity [63] Likert scale NOR1 I believe that it is fair and obligatory to help others because I know
that other people will help me some day
Dropped 0.8003
NOR2 I believe that other people will help me when I need help if I share
knowledge with others through MELANI
0.8464
NOR3 I believe that other people will answer my questions regarding specif-
ic information and knowledge in the future if I share knowledge with
others through MELANI
0.8714
NOR4 I think that people who are involved with MELANI develop recipro-
cal beliefs on give and take based on other people’s intentions and
behavior
0.6946
Executional cost [106] Likert scale EC1 I cannot seem to find the time to share knowledge in the community 0.6964 0.7882
EC2 It is laborious to share knowledge in the community 0.6950
EC3 It takes me too much time to share knowledge in the community 0.8626
EC4 The effort is high for me to share knowledge in the community 0.7913
Reputation [106] Likert scale R1 Sharing knowledge can enhance my reputation in the community 0.6312 0.6996
R2 I get praises from others by sharing knowledge in the community 0.6890
R3 I feel that knowledge sharing improves my status in the community 0.7922
R4 I can earn some feedback or rewards through knowledge sharing that
represent my reputation and status in the community
0.7039
Trust [87] Likert scale TR1 I believe that my colleague’s information security knowledge is
reliable
0.7510 0.8598
TR2 I believe that my colleague’s information security knowledge is
effective
0.8688
TR3 I believe that my colleague’s information security knowledge miti-
gates the risk of information security breaches
0.8460
TR4 I believe that my colleague’s information security knowledge is useful 0.8039
TR5 I believe that my colleagues would not take advantage of my infor-
mation security knowledge that we share
Dropped
6Journal of Cybersecurity, 2019, Vol. 5, No. 1
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
Results
Table 3 provides descriptive statistics for all variables. Table 4 speci-
fies Spearman correlations; for the sake of brevity, correlates for
controls are omitted. Tables 5 and 6document all models and their
respective diagnostic measures. Since we handled missing data con-
servativelyby list-wise deletion, the sample size of the respective
models is smaller than that of the full sample.
H1 is partially supported. A positive attitude towards SIS is posi-
tively associated with the intensity (P<0.05), but not with the fre-
quency of SIS. This may suggest that whenever the focal individual
believes SIS is an effective activity, his or her behavior is responsive
to information shared by other individuals.
H2a is fully supported. Social reciprocity is associated with both
the intensity (P<0.01) and the frequency of SIS (P<0.05). This
finding is in line with our theoretical expectation that individuals
seek equitable exchange relationships in which cooperative behavior
is rewarded. Future research may explore such social interaction
over time with a dyadic research setting, studying how exchange
patterns of repeated reciprocation develop over time.
H2b is partially supported. Transactional reciprocity is associ-
ated with the frequency of SIS (P<0.01), but not with its intensity.
This may imply that transactional rewards such as bonuses or pro-
motion motivate individuals to share knowledge they already pos-
sess with others in order to signal a high level of productive activity
vis-a`-vis their superiors.
H3 is fully supported. Consistent with our theoretical expect-
ation, executional cost is negatively associated with both the fre-
quency (P<0.05) and the intensity (P<0.001) of SIS. This not only
signals that executional cost constitutes a form of transaction cost
that may deter individuals from sharing, as we hypothesized. The
negative association with intensity is much stronger, suggesting that
the negative association of executional cost is larger when the focal
individual reacts to information shared by others. In other words, in
the presence of high executional cost, individuals seem to be pun-
ished for reacting. Since our research design only accounted for the
presence of executional cost, more research is required to identify
the institutional or organizational sources of this cost.
H4 is not supported. Contrary to what we hypothesized, we
find no support for the claim that an individuals’ expectation to in-
crease his or her status or social esteem is associated with SIS. Our
measure of reputation is neither significantly associated with the in-
tensity nor with the frequency of SIS. This negative result may be
due to the fact that Ref. 106 introduced their measure of reputation
(which we use in our empirical study) in the context of public know-
ledge sharing among private individuals who vie for public social es-
teem. In contrast, we study a population of security professionals in
the context of a private setting in which sensitive and classified in-
formation is shared. This may imply that, insofar as security infor-
mation sharing is concerned, future research should propose
alternative measures of reputation that are congruent with this
context.
Table 2: Final set of factor loadings after oblique rotation
a
Item Loading on oblimin-rotated factor Uniqueness
Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6 Factor 7
ISKS1 0.8075 0.27
ISKS2 0.8903 0.19
ISKS3 0.885 0.20
ISKS4 0.86 0.21
ISKS5 0.6898 0.44
AT2 0.7751 0.32
AT3 0.3412 0.6376 0.38
AT4 0.7849 0.31
NOR2 0.8464 0.23
NOR3 0.8714 0.18
NOR4 0.6946 0.36
HR1 0.8822 0.16
HR2 0.8743 0.19
HR4 0.7499 0.41
EC1 0.6964 0.49
EC2 0.695 0.45
EC3 0.8626 0.21
EC4 0.7913 0.32
R1 0.6312 0.49
R2 0.689 0.51
R3 0.7922 0.29
R4 0.7039 0.44
TR1 0.751 0.36
TR2 0.8688 0.21
TR3 0.846 0.26
TR4 0.8039 0.29
Eigenvalue 3.786 2.951 2.502 2.329 2.24 2.142 1.851
Proportion of
variance explained (%)
14.56 11.35 9.62 8.96 8.62 8.24 7.12
Cumulative
variance explained (%)
14.56 25.91 35.53 44.49 53.11 61.34 68.46
a
Blank cells represent factor loadings smaller than 0.30.
Journal of Cybersecurity, 2019, Vol. 5, No. 1 7
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
H5 is partially supported. The extent to which the focal individ-
ual trusts the information received will be useful is positively associ-
ated with the frequency (P<0.01), but not with the intensity of SIS.
This may imply that a focal individual who has such trust would be
more willing to share knowledge s/he already possesses. In this re-
spect, more research is required regarding the relationship between
initial trust among individuals and the evolution of such trust as ex-
change relationships unfold.
As regards the interaction effects, we find that H6a is partially
supported. The extent to which the focal individual trusts the infor-
mation received will be useful negatively moderates the relationship
between attitude and the intensity (P<0.05), but not the frequency
of SIS. This may imply that trust can function as a partial substitute
for attitude, in that the focal individual needs to convince him- or
herself to a lesser extent that SIS is useful in general if that individual
trusts the particular information s/he is about to receive is useful.
H6b is not supported. The extent to which the focal individual
trusts the information received will be useful neither moderates the
positive association of social reciprocity with the intensity of SIS nor
that with the frequency of SIS. This may imply that, unlike in the
above case for H6a, the focal individual’s trust that any particular
unit of information is useful cannot function as a substitute for the
importance of social reciprocity in the exchange relationship as
such.
H6c is fully supported. The extent to which the focal individual
trusts the counterparty provides valuable information negatively
moderates both the association of transactional reciprocity with the
frequency (P<0.01) and with the intensity (P<0.05) of SIS. In line
with our theoretical reasoning, this result may suggest that trust can
help the focal individual to convince him- or herself that the ex-
change relationship is equitable (since the information s/he is about
to receive is trusted to be useful), such that the focal individual has
to rely less on the expectation that s/he will be compensated by mon-
etary or career benefits whenever s/he participates in exchange
relationships.
Finally, the fact that we find partial support for H1, H2b, H5,
and H6a suggests that a differentiation of the theoretical construct
SIS into different measurement constructs is productive. Future re-
search may further develop the measures of frequency and intensity
we have proposed here or develop yet other detailed
operationalizations.
As regards our control variables, we find no significant associ-
ation of respondents’ demographic heterogeneity, length of member-
ship in MELANI-net, or industry affiliation with SIS. The latter
non-finding also alleviates concerns of overrepresentation of a par-
ticular industry or firm among the responses. For the controls “age,”
“industry,” and “education,” a benchmark category was automatic-
ally selected during estimation for every control (viz. footnotes b to
Tables 5 and 6).
The only significant association we find relates to the control
“education” in the model for the frequency of SIS. Since the educa-
tion category “other” is used as the benchmark, the results suggest
that in comparison to individuals with an education captured by
“other,” the remaining individuals in all other education categories
share significantly less in terms of frequency (P<0.01, respectively),
whereas no association with intensity is presented. Since all other
categories capture academic degrees and the case of no education,
this may imply that individuals who have a non-academic education
(e.g. vocational training) share knowledge they possess more often
with other individuals, probably because they are industry practi-
tioners who wish to propagate information they possess throughout
and across industries to strengthen organizational practice.
Discussion
Building on prior research in the field of the economics of informa-
tion security, and adopting a behavioral framework to organize our
theoretical reasoning, we have proposed how and why human be-
havior should be associated with SIS. To the best of our knowledge,
this study is the first that associates the self-reported sharing of sen-
sitive information among real individuals inside a private
Information Sharing and Analysis Center (ISAC) with the behavior
of these individuals. We also provide a dual empirical
Table 4: Correlations among dependent and independent variables
a
Frequency Intensity Attitude Reciprocity
(social)
Reciprocity
(transactional)
Executional
cost
Reputation Trust
Frequency 1
Intensity 0.3547*** 1
Attitude 0.2436*** 0.2742*** 1
Reciprocity (social) 0.2602*** 0.2750*** 0.3798*** 1
Reciprocity (transactional) 0.1836** 0.0456 0.0901 0.000 1
Executional cost 0.2238** 0.1694* 0.0976 0.0314 0.1533* 1
Reputation 0.0226 0.0968 0.1227 0.3069*** 0.0270 0.1148 1
Trust 0.2279** 0.0101 0.2471*** 0.0269*** 0.1321 0.1857* 0.1042 1
a
Spearman correlations.
*P<0.05; **P<0.01; ***P<0.001.
Table 3: Descriptive statistics on all variables
Variable Obs Mean SD Min Max
Frequency 240 2.68 0.78 1 5
Intensity 228 2.34 1.20 1 7
Attitude 208 4.10 0.53 3 5
Reciprocity (social) 195 3.88 0.60 1.66 5
Reciprocity (transactional) 195 2.16 0.75 1 4
Executional cost 208 3.14 0.65 1.25 5
Reputation 190 3.46 0.47 1.5 5
Trust 190 3.82 0.55 1.25 5
Gender 260 1.04 0.20 1 2
Age category 261 2.87 0.86 1 4
Education category 260 2.58 1.25 1 6
Membership duration 260 7.05 5.35 1 18
8Journal of Cybersecurity, 2019, Vol. 5, No. 1
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
operationalization of SIS by introducing the measures of SIS fre-
quency and intensity. Finally, our study confirms that interdisciplin-
ary approaches which attempt to integrate thinking from economics
and psychology are useful when SIS is studied [6].
Our study also contributes to prior work that has both theoretic-
ally predicted and descriptively noted that SIS, while beneficial, is
underutilized [16,32,33,44,47,72,73,103]. We provide some
first empirical evidence on the association of particular human
behaviors with SIS among individuals in a private ISAC setting. The
study also contributes to understanding the theoretical prediction
that actual SIS may not reach its societally optimal level [41,47]by
suggesting that human behavior may be at the core of this problem.
At the same time, we would caution regulators and researchers to
infer that SIS should be mandated (i.e. that individuals should be
forced to share) as a consequence of this problem. Adjusting sanc-
tion levels for failure to comply with mandatory SIS could be diffi-
cult, if not impossible [65]. Moreover, regulation that attempts to
solve the “sharing dilemma” in SIS should try to fix causes, not
symptoms [19]. Our study has collected cross-sectional data, and
hence we cannot establish causal relationships between human be-
havior and SIS. Nevertheless, the negative and significant associ-
ation between executional cost and both the frequency and intensity
of SIS that we identify confirms prior research that finds that institu-
tions shape human interaction and behavior. Institutions are formal
and informal rules which govern human behavior by rewarding de-
sirable actions and making undesirable actions more expensive or
punishable [12,75,76]. The organization of an ISAC is shaped by
both internal institutions (i.e. rules voluntarily agreed to among
ISAC participants and organizers) and external institutions (i.e. rules
imposed onto them by government and regulatory authorities).
Since high executional cost can be attributed to both effects, legisla-
tors, and regulators should be careful to predict the impact and con-
sequences of intended regulation for the executional cost of SIS. The
association between executional cost and SIS that our study identi-
fies suggests that humans are likely to assess the economic conse-
quences of external institutions in terms of executional costs and
adapt their behavior accordingly. Moreover, we find that both social
and transactional reciprocity are positively associated with both the
frequency and the intensity of SIS. Since reciprocity is a social norm,
it cannot be enforced by formal regulation and constraint, and the
attempt to do so may induce individuals to comply with the letter ra-
ther than the spirit of the law by sharing irrelevant, non-timely, or
false information [23].
We believe that the future study of these issues opens up prom-
ising paths for research that can both explain why individuals at-
tempt to circumvent SIS regulation and suggest more conducive
institutions. In this way, our study provides a stepping stone on
which future research can build. The extant literature has docu-
mented well that actual SIS, while considered highly useful in gen-
eral, is at low levels, and that individuals attempt to circumvent
regulation that makes SIS mandatory [5,32,33,71,72]. Our study
adds to these findings by suggesting that this economic problem of
Table 5: Models for intensity of SIS (ordered probit estimation)
a,b
Baseline Main effects Full model
Coefficient (robust standard error) Coefficient (robust standard error) Coefficient (robust standard error)
Attitude 0.4973 (0.1609)** 0.3627 (0.1672)*
Reciprocity (social) 0.3481 (0.1549)* 0.4045 (0.1526)**
Reciprocity (transactional) 0.2254 (0.1138)* 0.1860 (0.1118)
Executional cost 0.3949 (0.1198)*** 0.4833 (0.1314)***
Reputation 0.0083 (0.1905) 0.0932 (0.1895)
Trust 0.2250 (0.1577) 0.1847 (0.1501)
Attitude trust 0.6544 (0.2874)*
Reciprocity (social) trust 0.1969 (0.2431)
Reciprocity (transactional) trust 0.4561 (0.2119)*
Gender 0.2045 (0.3712) 0.1507 (0.4480) 0.2106 (0.4788)
Age 21–30 0.0920 (0.3434) 0.1286 (0.4063) 0.1361 (0.4204)
Age 31–40 0.0567 (0.2031) 0.0896 (0.2220) 0.1139 (0.2293)
Age 41–50 0.0001 (0.1762) 0.0138 (0.1777) 0.0096 (0.1820)
Education none 0.2253 (0.4789) 0.7976 (0.5208) 0.7239 (0.6388)
Education Master/Diploma 0.3512 (0.4649) 0.8990 (0.4964) 0.8336 (0.6368)
Education Bachelor 0.0206 (0.4635) 0.3347 (0.4924) 0.3198 (0.6202)
Education PhD 0.4581 (0.4984) 0.8959 (0.5322) 0.9997 (0.6382)
Membership duration 0.0257 (0.0134) 0.0184 (0.0165) 0.0184 (0.0164)
Government 0.1539 (0.2729) 0.2662 (0.3125) 0.2945 (0.3082)
Banking and Finance 0.0672 (0.2098) 0.1598 (0.2527) 0.1515 (0.2472)
All other industries 0.0472 (0.2473) 0.1649 (0.2977) 0.1576 (0.2982)
Energy 0.0283 (0.2931) 0.0650 (0.3260) 0.1007 (0.3217)
Health 0.3250 (0.2638) 0.2498 (0.3260) 0.2958 (0.3528)
Log pseudolikelihood 318.98 249.82 246.50
Pseudo R
2
0.0214 0.0773 0.0896
Wald v
2
(df) 16.10 (14) 55.43 (20)*** 64.02 (23)***
Observations 225 188 188
AIC kBIC 677.97 k746.29 551.65 k635.80 551.02 k644.87
a
Two-tailed tests.
b
Age category “above 50,” education category “other” and the telecommunication/IT industry serve as the respective control variable benchmarks.
*P<0.05; **P<0.01; ***P<0.001.
Journal of Cybersecurity, 2019, Vol. 5, No. 1 9
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
underutilization is difficult to resolve unless regulators and law-
makers consider the association of human behavior and SIS out-
comes. At this time, we speculate that a liberal institutional
environment that attempts to make individuals comply by
“nudging” them is probably more conducive than the attempt to
enforce compliance by coercion [91]. We leave it to future research
to either corroborate or refute this speculation, suggesting that ir-
respective of any particular institutional arrangement, human be-
havior is significantly associated with SIS and hence likely
responds to changes in institutional configuration. All in all, our
study suggests that future research can productively employ behav-
ioral theory and methods as it attempts to further develop SIS re-
search by considering the human interaction that precedes actual
acts of sharing.
In a broader sense, our work develops prior conceptual ideas
that human aspects matter at least as much as technological ones
when SIS is concerned [19]. Our empirical approach takes the
technological context as a given and focuses on identifying associa-
tions between human behavior and SIS. Cybersecurity managers in
organizations can benefit from these results as they attempt to make
individuals comply with organizational goals. Our results suggest
that both the frequency and the intensity of SIS are associated with
human behavior. Managers should therefore be careful to study
these associations when they define organizational goals and accept
that individual human behavior does not necessarily comply with
these unless appropriate goal alignment is provided [57,66]. For ex-
ample, managers may facilitate an individual’s participation in SIS
by reducing the executional cost of information exchange, or they
may provide the focal individual with intelligence on counterparties
to help them assess the likelihood with which information sharing
may be reciprocated.
Our study is pioneering in the sense that it studies real human
beings and their self-reported behavior in the context of a real ISAC.
Nevertheless, it merely studies a single, centrally organized ISAC in
a single country. Hence, future research should generalize our ap-
proach to alternative models of ISAC organization and explore di-
verse national and cultural settings by replicating our study with
different ISACs and nation-states. We believe our approach is con-
ducive to such generalization since neither our theoretical frame-
work, nor any one of our behavioral constructs, nor the empirical
measures we used to operationalize these are context-specific to any
particular national or cultural context. Our measures and the theory
in which they are grounded rather represent fundamental aspects of
human behavior which, in our view, should apply globally. Thus, fu-
ture work could complement our study with data from different
ISACs, such that a transnational continuum of sharing intensities
and frequencies could be constructed. This continuum would allow
researchers to identify commonalities and differences in information
exchange patterns and use these insights to propose expedient policy
options.
Table 6: Models for frequency of SIS (Tobit estimation)
a,b
Baseline Main effects Full model
Coefficient (robust standard error) Coefficient (robust standard error) Coefficient (robust standard error)
Attitude 0.2797 (0.1214)* 0.1895 (0.1111)
Reciprocity (social) 0.1807 (0.1195) 0.2150 (0.1046)*
Reciprocity (transactional) 0.2734 (0.0824)** 0.2361 (0.0816)**
Executional cost 0.1872 (0.0911)* 0.2336 (0.0962)*
Reputation 0.1827 (0.1243) 0.1121 (0.1232)
Trust 0.2689 (0.1058)* 0.2964 (0.1036)**
Attitude trust 0.3490 (0.2311)
Reciprocity (social) trust 0.2055 (0.1813)
Reciprocity (transactional) trust 0.3839 (0.1378)**
Gender 0.4851 (0.1681)** 0.2412 (0.1791) 0.1837 (0.1773)
Age 21–30 0.1852 (0.2131) 0.2595 (0.2387) 0.2057 (0.2378)
Age 31–40 0.0365 (0.1567) 0.0218 (0.1528) 0.0051 (0.1513)
Age 41–50 0.0294 (0.1222) 0.0040 (0.1264) 0.0171 (0.1243)
Education none 0.6274 (0.1705)*** 0.9126 (0.2210)*** 0.8152 (0.2441)**
Education Master/Diploma 0.6063 (0.1462)*** 0.8749 (0.2291)*** 0.7984 (0.2671)**
Education Bachelor 0.5872 (0.1531)*** 0.8089 (0.2062)*** 0.7678 (0.2324)**
Education PhD 0.5392 (0.2667)* 0.8892 (0.2976)** 0.9345 (0.3181)**
Membership duration 0.0277 (0.0112)* 0.0211 (0.0118) 0.0213 (0.0112)
Government 0.1629 (0.2039) 0.0130 (0.2373) 0.0097 (0.2288)
Banking and Finance 0.0613 (0.1694) 0.0328 (0.2142) 0.0304 (0.2064)
All other industries 0.5292 (0.1947)** 0.4016 (0.2430) 0.3748 (0.2395)
Energy 0.1054 (0.2236) 0.2191 (0.2485) 0.1867 (0.2399)
Health 0.0909 (0.2115) 0.0767 (0.2787) 0.0465 (0.2759)
Constant 2.6652 (0.2705)*** 3.0954 (0.3520)*** 3.0939 (0.3577)***
Log pseudolikelihood 274.73 202.46 197.92
Pseudo R
2
0.0538 0.1370 0.1564
F(df) 4.10 (14, 223)*** 5.25 (20, 168)*** 5.25 (23, 165)***
Observations (left kright censored) 237 (12 k1) 188 (10 k1) 188 (10 k1)
AIC kBIC 581.47 k636.96 448.93 k520.13 445.84 k526.75
a
Two-tailed tests.
b
Age category “above 50,” education category “other” and the telecommunication/IT industry serve as the respective control variable benchmarks.
*P<0.05; **P<0.01; ***P<0.001.
10 Journal of Cybersecurity, 2019, Vol. 5, No. 1
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
Finally, the ISACs that exist as of today have evolved from trade
associations, government agencies, and public–private partnerships.
However, the evolution of such historical trajectories is subject to
technological change [74]. We therefore believe that novel technolo-
gies could facilitate human interaction in future ISAC configura-
tions. For example, since the cost of reputation losses upon security
breaches can be interpreted as privacy risk [19], insights from priv-
acy research and secure distributed computation and interaction
[35] might be used to construct distributed ISACs with safe real-
time participation. Future research may use our study to consider
the impact of such novel technological approaches on human behav-
ior to prevent unintended consequences.
From a broader perspective, our study design has some limita-
tions that point to opportunities for future research.
4
First, both as
regards the level and the unit of analysis, our study focuses on the in-
dividual. This implies that interactions between the individual and
the organizational and institutional contexts within which the focal
individual acts are beyond the scope of this study. Nevertheless, our
setting may be expanded both theoretically and empirically to
incorporate such multilevel interactions. For example, the
organizational-level performance implications of SIS could be
studied, in that future research would analyze the association of in-
dividual behavior with organizational results, such as increased
cybersecurity or increased financial performance.
In particular, future research may analyze the extent to which
different organizational processes, cultures, and risk management
approaches are associated with SIS by way of human behavior. For
example, critical infrastructure providers who face significant risks
of business interruption and going concern if their cybersecurity is
compromised may emphasize more than other organizations that
SIS is desirable and hence direct their employees to act accordingly.
Thus, organizational policy may moderate the association between
human behavior and SIS. Future research could build on our ap-
proach by developing more complex multilevel study designs that
can incorporate such additional sources of variance.
Finally, our study design is cross-sectional, implying that we can
only claim association, but not causation. While we believe this is
acceptable given the pioneering nature of this study, controlled
experiments are required to establish causality. We encourage future
work to introduce such methods. Further, future studies could also
ethnographically analyze human interaction within an ISAC over
time, log how and why behavior changes, and infer how this behav-
ioral evolution operates on SIS outcomes.
Conflict of interest statement. None declared.
Acknowledgments
Comments received from Tyler Moore, Thomas Maillart, Mauro Cherubini,
and three anonymous reviewers all helped to improve previous versions of
this article. We thank the Swiss Federal Institute of Technology Zurich for
providing free software and IT support during the survey implementation
phase. The usual disclaimer applies.
References
1. Ajzen I. The directive influence of attitudes on behavior. In: Gollwitzer PM
and Bargh JA (eds), The Psychology of Action: Linking Cognition and
Motivation to Behavior. New York, NY: Guilford Press, 1996, 385–403.
2. Ajzen I, Fishbein M. Understanding Attitudes and Predicting Social
Behavior. Englewood Cliffs, NJ: Prentice-Hall, 1980.
3. Ajzen I, Madden TJ. Prediction of goal-directed behavior: attitudes,
intentions, and perceived behavioral control. J Exp Soc Psychol 1986;22:
453–74.
4. Amayah AT. Determinants of knowledge sharing in a public sector or-
ganization. J Knowledge Manage 2013;17:454–71.
5. Anderson R, Fuloria S. Security economics and critical national infra-
structure. In: Moore T, Pym D, Ioannidis C (eds), Economics of
Information Security and Privacy. Boston, MA: Springer, 2010, 55–66.
6. Anderson R, Moore T. The economics of information security. Science
2006;314:610–13.
7. Andreoni J. Cooperation in public-goods experiments: kindness or con-
fusion? Am Econ Rev 1995;85:891–904.
8. Aviram A, Tor A. Overcoming impediments to information sharing. Ala
Law Rev 2003;55:231–80.
9. Ba S, Pavlou P. Evidence of the effect of trust building technology in elec-
tronic markets: price premiums and buyer behavior. MIS Quart 2002;
26:243–68.
10. Bauer J, van Eeten M. Cybersecurity: stakeholder incentives, external-
ities, and policy options. Telecommun Policy 2009;33:706–19.
11. Baumeister RF, Leary MR. The need to belong: desire for interpersonal
attachments as a fundamental human motivation. Psychol Bull 1995;
117:497–529.
12. Baumol WJ. Entrepreneurship: productive, unproductive, and destruc-
tive. J Political Econ 1990;98:893–921.
13. Bazerman MH. Judgement in Managerial Decision Making. New York,
NY: Wiley, 2005.
14. Belletier C, Robert A, Motak L, et al. Toward explicit measures of inten-
tion to predict information system use: an exploratory study of the role
of implicit attitudes. Comput Human Behav 2018;86:61–8.
15. Be´ nabou R, Tirole J. Incentives and prosocial behavior. Am Econ Rev
2006;96:1652–78.
16. Bisogni F. Data breaches and the dilemmas in notifying customers. In:
Workshop on the Economics of Information Security (WEIS), Delft,
2015.
17. Bock GW, Zmud RW, Kim YG, et al. Behavioral intention formation in
knowledge sharing: examining the roles of extrinsic motivators, social–
psychological forces, and organizational climate. MIS Quart 2005;29:
87–112.
18. Bodin LD, Gordon LA, Loeb MP, et al. Cybersecurity insurance and
risk-sharing. J Account Pub Policy 2018;37:527–44.
19. Bo¨ hme R. Back to the roots: information sharing economics and what
we can learn for security. In: Second Workshop on Information Sharing
and Collaborative Security (WISCS), Denver, CO: ACM, 2015.
20. Bolton GE, Ockenfels A. ERC: a theory of equity, reciprocity, and com-
petition. Am Econ Rev 2000;90:166–93.
21. Brennan G, Pettit P. The Economy of Esteem. Oxford: Oxford
University Press, 2004.
22. Brosnan SF, de Waal F. Monkeys reject unequal pay. Nature 2003;425:
297–99.
23. Burr R. To Improve Cybersecurity in the United States through
Enhanced Sharing of Information about Cybersecurity Threats, and for
Other Purposes. Washington, DC: 114th United States Congress, 2015.
24. Chaiken S. Heuristic versus systematic information processing and the
use of source versus message cues in persuasion. J Pers Soc Psychol 1980;
39:752–66.
25. Chang HH, Chuang S-S. Social capital and individual motivations on
knowledge sharing: participant involvement as a moderator. Inf Manage
2011;48:9–18.
26. Cohen J, Cohen P, West SG, et al.Applied Multiple Regression/
Correlation Analysis for the Behavioral Sciences. 3rd ed. London: Taylor
& Francis, 2002.
4 We thank two anonymous reviewers for sharing ideas about how our ap-
proach may be expanded and generalized.
Journal of Cybersecurity, 2019, Vol. 5, No. 1 11
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
27. DellaVigna S. Psychology and economics: evidence from the field. J Econ
Lit 2009;47:315–72.
28. Dillman DA, Smyth J, Christian LM. Internet, Phone, Mail, and Mixed-
Mode Surveys: The Tailored Design Method, 4th edn. Hoboken, New
Jersey: John Wiley & Sons, 2014.
29. Dunn Cavelty M. Cybersecurity in Switzerland. Springer Briefs in
Cybersecurity. Cham: Springer International Publishing, 2014.
30. Eagly AH, Chaiken S. The Psychology of Attitudes. Fort Worth, TX:
Harcourt et al., 1993.
31. Emler N. A social psychology of reputation. Eur Rev Soc Psychol 1990;
1:171–93.
32. ENISA. Incentives and Barriers to Information Sharing. Heraklion:
European Union Agency for Network and Information Security, 2010.
33. ENISA. Information Sharing and Common Taxonomies between
CSIRTs and Law Enforcement. Heraklion: European Union Agency for
Network and Information Security, 2016.
34. ENISA. Information Sharing and Analysis Centres (ISACs). Cooperative
Models. Heraklion: European Union Agency for Network and
Information Security, 2017.
35. Ezhei M, Ladani BT. Information sharing vs. privacy: a game theoretic
analysis. Expert Syst Appl 2017;88:327–37.
36. Fehr E, Ga¨ chter S. Fairness and retaliation: the economics of reciprocity.
J Econ Perspect 2000; 14:159–81.
37. Fehr E, Ga¨ chter S. Altruistic punishment in humans. Nature 2002;415:
137–40.
38. Fehr E, Gintis H. Human motivation and social cooperation: experimen-
tal and analytical foundations. Annu Rev Sociol 2007;33:43–64.
39. Fehr E, Schmidt K. A theory of fairness, competition, and cooperation.
Quart J Econ 1999;114:817–68.
40. Gabaix X, Laibson D, Moloche G, et al. Costly information acquisition:
experimental analysis of a boundedly rational model. Am Econ Rev
2006;96:1043–68.
41. Gal-Or E, Ghose A. The economic incentives for sharing security infor-
mation. Inf Syst Res 2005;16:186–208.
42. Gal-Or E, Ghose A. The economic consequences of sharing security in-
formation. In: Camp LJ, and Lewis S (eds), Economics of Information
Security. Advances in Information Security, Vol. 12. Boston, MA:
Springer, 2004.
43. Gefen D, Karahanna E, Straub DW. Trust and TAM in online shopping:
an integrated model. MIS Quart 2003;27:51–90.
44. Ghose A, Hausken K. A strategic analysis of information sharing among
cyber attackers. SSRN Electron J 2006;12:1–37.
45. Gordon LA, Loeb MP, Lucyshyn W, et al. Externalities and the magni-
tude of cyber security underinvestment by private sector firms: a modifi-
cation of the Gordon-Loeb model. J Inf Secur 2015;6:24–30.
46. Gordon LA, Loeb MP, Lucyshyn W, et al. The impact of information
sharing on cybersecurity underinvestment: a real options perspective. J
Account Public Policy 2015;34:509–519.
47. Gordon LA, Loeb MP, Lucyshyn W. Sharing information on computer
systems security: an economic analysis. J Account Public Policy 2003;22:
461–85.
48. Gordon LA, Loeb MP, Sohail T. Market value of voluntary disclosures
concerning information security. MIS Quart 2010;34:567–94.
49. Gordon LA, Loeb MP, Zhou L. Investing in cybersecurity: insights from
the Gordon-Loeb model. J Inf Secur 2016;7:49.
50. Gouldner AW. The norm of reciprocity: a preliminary statement. Am
Sociol Rev 1960;25:161–78.
51. Granovetter M. Economic action and social structure: the problem of
embeddedness. Am J Sociol 1985;91:481–510.
52. Hair JF, Black WC, Babin BJ, et al.Multivariate Data Analysis, 5th edn.
Upper Saddle River, NJ: Prentice Hall, 2009.
53. Harrison K, White G. Information sharing requirements and framework
needed for community cyber incident detection and response. In: 2012
IEEE Conference on Technologies for Homeland Security (HST),
Waltham, IEEE, 2012, 463–69.
54. Hausken K. A strategic analysis of information sharing among cyber
attackers. J Inf Syst Technol Manage 2015;12:245–70.
55. Hausken K. Information sharing among firms and cyber attacks. J
Account Public Policy 2007;26:639–88.
56. Hsu M-H, Ju TL, Yen C-H, et al. Knowledge sharing behavior in virtual
communities: the relationship between trust, self-efficacy, and outcome
expectations. Int J Human-Comput Stud 2007;65:153–69.
57. Hume D. A Treatise of Human Nature. New York, NY, Oxford
University Press, 2000.
58. Kahneman D, Tversky A. Prospect theory: an analysis of decision under
risk. Econometrica 1979;47:263–91.
59. Kahneman D, Tversky A. Prospect theory—an analysis of decision under
risk. Econometrica 1979;47:263–91.
60. Keith B, Babchuk N. The quest for institutional recognition: a longitu-
dinal analysis of scholarly productivity and academic prestige among
sociology departments. Social Forces 1998;76:1495–1533.
61. Kolm S-C, Ythier JM. Handbook of the Economics of Giving, Altruism
and Reciprocity. Amsterdam: Elsevier, 2006.
62. Kroenung J, Eckhardt A. The attitude cube—a three-dimensional model
of situational factors in IS adoption and their impact on the attitude–
behavior relationship. Inf Manage 2015;52:611–27.
63. Kwahk K-Y, Park D-H. The effects of network sharing on knowledge-
sharing activities and job performance in enterprise social media environ-
ments. Comput Human Behav 2016;55:826–39.
64. Laube S, Bo¨ hme R. Strategic aspects of cyber risk information sharing.
ACM Comput Surv (CSUR) 2017;50:1–77.
65. Laube S, Bo¨hme R. The economics of mandatory security breach report-
ing to authorities. J Cybersecur 2016;2:29–41.
66. Lindenberg S, Foss N. Managing joint production motivation: the role of goal
framing and governance mechanisms. Acad Manage Rev 2011;36:500–25.
67. Luiijf E, Klaver M. On the sharing of cyber security information. In: Rice
M, Shenoi S (eds), Critical Infrastructure Protection IX. Cham: Springer,
2015, 29–46.
68. Malhotra D. Trust and reciprocity decisions: the differing perspectives of
trustors and trusted parties. Org Behav Human Decis Process 2004;94:
61–73.
69. McElreath R. Reputation and the evolution of conflict. J Theor Biol
2003;220:345–57.
70. McEvily B, Perrone V, Zaheer A. Trust as an organizing principle. Org
Sci 2003;14:91–103.
71. Moore T. The economics of cybersecurity: principles and policy options.
Int J Crit Infrastruct Prot 2010;3:103–17.
72. Moran T, Moore T. The Phish-Market protocol: securely sharing attack
data between competitors. In: Sion R (ed.), Financial Cryptography and
Data Security. Berlin-Heidelberg: Springer, 2010, 222–37.
73. Naghizadeh P, Liu M. Inter-temporal incentives in security information
sharing agreements. In: 2016 Information Theory and Applications
Workshop (ITA). IEEE, 2016, 1–8.
74. Nelson RR, Winter SG. An Evolutionary Theory of Economic Change.
Cambridge: Belknap Press, 1982.
75. North DC. Institutions, Institutional Change and Economic
Performance. Cambridge: Cambridge University Press, 1990.
76. North DC. Understanding the Process of Economic Change. Cambridge:
Cambridge University Press, 2005.
77. Nunnally JC, Bernstein I. 2017. Psychometric Theory, 3rd edn. New
York: McGraw-Hill.
78. Oliver P. Rewards and punishments as selective incentives for collective
action: theoretical investigations. Am J Sociol 1980;85:1356–75.
79. Olson M. The Logic of Collective Action. Cambridge, MA: Harvard
University Press, 1965.
80. Paese PW, Gilin DA. When an adversary is caught telling the truth: recip-
rocal cooperation versus self-interest in distributive bargaining. Pers Soc
Psychol Bull 2000;26:79–90.
81. Park JH, Gu B, Leung ACM, et al. An investigation of information shar-
ing and seeking behaviors in online investment communities. Comput
Human Behav 2014;31:1–12.
12 Journal of Cybersecurity, 2019, Vol. 5, No. 1
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
82. Petty RE, Cacioppo, JT. The elaboration likelihood model of persuasion.
Adv Exp Soc Psychol 1986;19:123–205.
83. Podsakoff PM, MacKenzie S, Podsakoff NP. Sources of method bias in
social science research and recommendations on how to control it. Annu
Rev Psychol 2012;63:539–69.
84. Podsakoff PM, Organ DW. Self-reports in organizational research: prob-
lems and prospects. J Manage 1986;12:531–44.
85. Reinholt MIA, Pedersen T, Foss NJ. Why a central network position isn’t
enough: the role of motivation and ability for knowledge sharing in em-
ployee networks. Acad Manage J 2011;54:1277–97.
86. Ridings CM, Gefen D, Arinze B. Some antecedents and effects of trust in
virtual communities. Strategic Inf Syst 2002;11:271–95.
87. Safa NS, von Solms R. An information security knowledge sharing model
in organizations. Comput Human Behav 2016;57:442–51.
88. Siegrist J, Starke D, Chandola T, et al. The measurement of effort–
reward imbalance at work: European comparisons. Soc Sci Med 2004;
58:1483–99.
89. Simon HA. Administrative Behavior: A Study of Decision-Making Processes
in Administrative Organization. New York, NY: Free Press, 1976.
90. Smith EA, Winterhalder B. (eds) Evolutionary Ecology and Human
Behavior. New York, NY: Routledge, 2017.
91. Thaler RH, Sunstein CR. Nudge: Improving Decisions about Health,
Wealth, and Happiness. New York, NY: Penguin Books, 2009.
92. Tom S, Fox C, Trepel C, et al. The neural basis of loss aversion in
decision-making under risk. Science 2007;315:515–518.
93. Tomasello M, Carpenter M, Call J, et al. Understanding and sharing
intentions: the origins of cultural cognition. Behav Brain Sci 2005;28:
675–91.
94. Trevor CO, Nyberg AJ. Keeping your headcount when all about you are
losing theirs: downsizing, voluntary turnover rates, and the moderating
role of HR practices. Acad Manage J 2008;51:259–76.
95. Tricomi E, Rangel A, Camerer C, et al. Neural evidence for inequality-
averse social preferences. Nature 2010;463:1089–1091.
96. Tversky A, Kahneman D. Loss aversion in riskless choice: a reference de-
pendent model. Quart J Econ 1991;106:1039–61.
97. Tversky A, Kahneman D. Advances in prospect theory: cumulative repre-
sentation of uncertainty. J Risk Uncertainty 1992;5:297–323.
98. Vakilinia I, Louis SJ, Sengupta S. Evolving sharing strategies in cyberse-
curity information exchange framework. In: Proceedings of the Genetic
and Evolutionary Computation Conference Companion. New York,
NY: ACM, 2017, 309–10.
99. von Hippel E, von Krogh G. Open source software and the “private-
collective” innovation model: issues for organization science. Org Sci
2003;14:209–23.
100. Wang W-T, Hou Y-P. Motivations of employees’ knowledge sharing
behaviors: a self-determination perspective. Inf Org 2015;25:1–26.
101. Wang JH, Wang C, Yang J, et al. A study on key strategies in P2P file
sharing systems and ISPs’ P2P traffic management. Peer-to-Peer
Network Appl 2011;4:410–19.
102. Watzlawick P, Bavelas JB, Jackson DD. Pragmatics of Human
Communication, New York, Norton & Company, 2011.
103. Weiss E. Legislation to Facilitate Cybersecurity Information Sharing:
Economic Analysis. Washington, DC: Congressional Research Service,
2015.
104. Williamson OE. The economics of organization: the transaction cost ap-
proach. Am J Sociol 1981;87:548–77.
105. Xiong L, Liu L. PeerTrust: supporting reputation-based trust for peer-to-
peer electronic communities. IEEE Trans Knowledge Data Eng 2004;16:
843–57.
106. Yan Z, Wang T, Chen Y, et al. Knowledge sharing in online health com-
munities: a social exchange theory perspective. Inf Manage 2016;53:
643–53.
107. Zhao W, White G. A collaborative information sharing framework
for Community Cyber Security. In: IEEE Conference on Technologies
for Homeland Security (HST), Waltham, MA: IEEE, 2012, 457–62.
Journal of Cybersecurity, 2019, Vol. 5, No. 1 13
Downloaded from https://academic.oup.com/cybersecurity/article-abstract/5/1/tyz006/5554880 by guest on 27 August 2019
