Determining the Value of Information Security Investments
A Decision Support System
Hannah Louise Davies and Andrew J. C. Blyth
Information Security Research Group, University of South Wales, Pontypridd, U.K.
Keywords: Return on Security Investment, Multi-Attribute Utility Theory, Decision Making under Uncertainty, Multi-
Criteria Decision Making, Information Security Metrics, Decision Making for Information Security.
Abstract: Advances in the technological era are making information security breaches a more common occurrence. A
vital part of ensuring an organisation is well protected from these increasingly complex threats is a suitable
security solution. Suitability of a security solution should not only be measured in terms of goals such as
reducing down time or reducing the risk of a certain threat, but also meet stakeholder and executive goals in
terms of being cost effective. Currently, cost effective is determined by calculating a return on security
investment calculation, where the cost of a solution is evaluated against any savings resulting after
purchasing the solution to determine whether the option is viable. The current implementation of return on
security investment calculations however is often subjective and inaccurate as calculations are performed in
an ad-hoc manner. When there are multiple factors to consider, with uncertain or incomplete values
available, a multi-attribute decision making method that utilises uncertainty is required in order to allow the
decision maker to assess all possible options in the most logical and objective manner, whilst keeping in
mind the goals of the organisation. In this paper we present and evaluate a conceptual, analytical framework
that, with the use of multi-attribute utility theory under uncertainty, is able to model return on security
investment calculations in a novel way. This new calculation is introduced as a Value of Information
Security Investment calculation. The final goal is to create a framework that allows for repeatable,
predictable and mature, calculations that determine the value of an information security investment.
1 INTRODUCTION
Since the global economic crisis of 2008, justifying
a decision to acquire tools, techniques and processes
has become increasingly important. In this
technology era, organisations are faced with building
and maintaining an effective information technology
infrastructure that is not only suited to the
organisations objectives, but conforms to policy,
addresses the alarming increase in information
security threats (Al-Humaigani and Dunn, 2003) and
is cost-effective.
When designing an effective security solution, an
organisation requires a complete Risk Management
framework to identify, measure and manage the
risks present. Assessing the financial aspect of a
security solution is also important and is most
commonly evaluated by performing a return on
investment (ROI) calculation (Pontes et al., 2011).
ROI calculations aim to answer the question as to
which option gives the most value for money
(Sonnenreich, 2006). The problem faced, however,
is that security programs and solutions are often
preventing losses and not delivery profit; hence
calculating a standard ROI is not sufficient.
Applying a security program or solution is expected
to reduce the risks threatening your assets, thus a
return on security investment (ROSI) calculation
should be used to calculate how much loss has been
avoided following the investment (ENISA, 2012).
The standard ROSI methodology involves the
comparison of the monetary value of the investment
and the monetary value of the risk reduction. The
monetary value of the risk reduction can be
estimated by a quantitative risk assessment (ENISA,
2012).
With security threats becoming more
sophisticated and more costly, determining the
acceptable risk level and selection of an optimal cost
effective solution is becoming an increasingly
complex but necessary decision for information
security management. As with all such decisions,
expenditure to protect a system has to be fully
justified in terms of a well-specified objective
(Ioannidis et al., 2010). Another issue that arises
426
Louise Davies H. and J. C. Blyth A..
Determining the Value of Information Security Investments - A Decision Support System.
DOI: 10.5220/0005170704260433
In Proceedings of the International Conference on Knowledge Management and Information Sharing (KMIS-2014), pages 426-433
ISBN: 978-989-758-050-5
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
with calculating ROSI is the lack of data concerning
information security (Levy, 2005). Security breaches
and incidents are still seen as a reason for
embarrassment and few companies are willing or
able to provide actual costs associated with security
breaches (Sonnenreich, 2006). However, whilst the
data itself can be improved in terms of accuracy, it is
in fact the reproducible nature of the calculation that
is important, this reproducible nature arises from a
consistent methodology of calculating and reporting
the cost (Sonnenreich, 2006).
There are multiple factors and numerous
alternatives to consider when such a widely used
methodology as multi-criteria decision-making
(MCDM) is used. MCDM is a collection of
methodologies used to compare, select or rank
multiple alternatives that typically involve
incommensurate attributes (Levy, 2005). MCDM
can fall in to two main categories; 1) decision
making under certainty and 2) decision making
under uncertainty. However, in practice, it is
extremely unlikely to have complete information
about the future (Shah et al., 2007). The values used
in ROSI calculations are subject to uncertainty,
primarily down to the fact that you cannot predict,
with a high level of certainty, the losses caused
when an event may or may not happen
(Sonnenreich, 2006). Consequently, decision-
making under uncertainty is the logical method to
use in order to allow for realistic solutions.
Though there are previous applications of
decision analysis to ROSI problems, the inclusion of
multi-attribute utility theory (MAUT) hasn’t been
applied to its fullest extent. The cases of MAUT
being used are restricted to a small number of
attributes, such as cost and availability (Ioannidis et
al., 2010), (Beautement et al., 2008), or cost,
investment and availability (Beres, Pym and Shiu,
2010) for example. The limited use of MAUT
doesn’t show the full extent to which decision
analysis can be used to support ROSI calculations.
2 RELATED WORK
2.1 Return on Security Investment
Performing an ROSI calculation is a method of
evaluating a security investment prior to making a
decision; it compares the cost of implementing and
procuring a solution to the losses avoided. It is the
misconception of what ROSI demonstrates, in terms
of not necessarily representing a profit, which has
led to its misuse.
Relatively recently, in 2002, Gordon and Loeb
adopted a static optimisation model where the
optimal ratio of investment in Information Security
can be calculated under different assumptions of
expected loss. This model relies on restrictive
assumptions to calculate the optimal ratio and has
sparked much debate regarding whether the
relationship between Information Security
Investment and vulnerability is always a monotonic
function. In 2006, Hausken proposed that
vulnerability be represented as a function, showing
the optimal ratio cannot be supported.
An important issue is that the inputs for ROSI
may be highly subjective, and consequently,
companies that use the same method for calculation
can arrive at extremely varied results due to different
choices made about the inclusion or exclusion of
costs (Sonnenreich, 2006). In addition, estimating
losses from future events brings uncertainty in to the
values used. Even when actuarial tables or insurance
data are used, these values may not be accurate due
to the “ostrich effect” (Sonnenreich, 2006)
experienced when an incident occurs. Another way
to collet data is to base values on competitors
experiences. This could be to investigate losses
incurred by competitors during incidents, or an
uptake in sales within an organisation once a
competitor has experienced an attack. These values
of losses or gains can be added on to the ROSI
analysis (Korostoff, 2003) or values based on
previous experience can also be used. Finally, the
variable most complex to define and evaluate is the
mitigated risk. One method to consider is the use of
past data, determining the expected losses due to
security breaches prior, and subsequent to,
implementing a solution (Arora, Frank and Telang,
2008). The avoided risk or expected benefit is then
the difference between the baseline loss and residual
risk. When using this approach, it must be
recognised that rare events may preclude the use of
past data, when such events occurred sufficiently
long ago that the same conditions no longer apply.
ROSI calculations, as explained above are based
heavily on estimations or perceptions of values –
this makes ROSI more of an approximation and less
accurate. It is the ability to manipulate these
approximations to justify decisions (ENISA, 2012)
that call for ROSI to be improved, in terms of
reproducibility, repeatability and predictability. The
biased perceptions of a decision maker should not
cause differing outputs of the calculations. ROSI
equations should be objective and numerical but due
to insufficient definition or subjective variables, the
calculations become imprecise and subjective
DeterminingtheValueofInformationSecurityInvestments-ADecisionSupportSystem
427
(Rudolph and Schwarz, 2012). Another criticism of
ROSI is that the methods are complex and rely on
soft data to derive hard numbers that are associated
with ROSI. This takes time and therefore simple but
effective methods are preferred in practice
(Holoman and Kuzmeskus, 2012). This research
aims to identify a more suitable method of justifying
Information Security Expenditure by means of an
accurate, reproducible and objective approach in
order to mitigate such criticism.
2.2 Multi-Attribute Utility Theory
Concerning ROSI, most decisions are made in an
environment where the eventual outcomes are
unknown. It is the area of decision making under
uncertainty where the methodology of ROSI and
decision analysis should be performed. Decision
analysis was originally developed to aid a decision-
maker’s evaluation of alternatives whose outcomes
were uncertain (Raïffa and Schlaifer, 1961).
Decision theory provides an approach to modelling
uncertainty in the environment, using probability to
describe the likelihood of uncertain events and using
utility to model the decision-maker’s attitude to risk
(Belton and Stewart, 2002).
Multi-attribute utility theory seeks to reduce the
complex problem of assessing a multi-attribute
utility function into one of assessing a series of uni-
dimensional utility functions. These individually
estimated component functions are then brought
together again; the glue is known as value trade-offs
(Zeleny, 1982). The trade-off issue is, in essence,
evaluating how much achievement, in objective 1, is
the decision maker willing to give up to improve
achievement, by some fixed amount, on objective 2
(Keeney and Raïffa, 1976). By determining the
trade-offs that a decision-maker is willing to make,
the optimal solution based on these preferences can
be determined.
An important step in using MAUT is defining the
objectives that an organisation has. There are 4
identified steps to develop and quantify objectives
(Keeney, 1975):
Identify objectives that the organisation
considers important.
Structure the objectives into a hierarchy so
ends and means remain distinct and
redundancies get eliminated.
For clarification, define attributes for
objectives, and then consequences are
measurable.
Develop a utility function to indicate value
trade-offs and reflect different viewpoints
within the organisation.
For the application of MAUT that is being
considered here, an objective may be to reduce the
vulnerability of the organisation to a specific threat
agent. An attribute would then be the number of
incidents in a week/month/year relating to that
specific threat agent. Uncertainty would then be
propagated over the variables with the use of
probabilities.
The basis of MAUT is the use of utility
functions; these can be applied to transform the raw
performance values of the alternatives against
diverse criteria, factual and judgemental, to a
common dimensionless scale (Fülöp, 2005). In
practice, the interval [0,1] is used, where 1 is the
highest utility that can be achieved, and 0 is the
lowest utility that can be achieved. Utility functions
allow for the most preferred alternative to obtain a
higher utility value than less preferred alternatives.
The use of utility functions also allow for more
relevant criteria to contribute more substantially to
the overall utility value than less relevant criteria.
The actual magnitude of the utility is not important,
the utility function serves as a “risk preference
thermometer” that can be used for ranking lotteries
according to the risk preference of an individual
(Howard, 1968). We seek a utility function U(z),
such that alternative a is preferred to alternative b if
and only if the expectation of U(z
a
) is greater than
U(z
b
), that is:
E[U(z
a
)] > E[U(z
b
)] (1)
Where z
a
and z
b
are the random vectors of attribute
values associated with alternatives a and b
respectively. This is the Expected Utility Hypothesis
(Løken, 2007). Establishing the existence of a utility
function is a non-trivial process, even for the single
attribute case. It is within this context that von
Neumann and Morgenstern (1947) first developed
utility theory.
2.2.1 Additive Utility Model
The additive utility function is the best-known form
of multi-attribute utility functions because of its
relative simplicity and application to real world
problems (Raïffa and Schlaifer, 1961). The additive
utility function should be considered firstly as a
possible solution as it is the simplest form of a utility
function. There are a number of conditions that need
to be shown to be satisfied in order for the additive
form of multi-attribute utility model to be valid.
However, when it is not possible to prove all of the
additivity assumptions hold for every case, strong
evidence suggests that even when complete
independence isn’t established, the additive model
KMIS2014-InternationalConferenceonKnowledgeManagementandInformationSharing
428
can still provide close approximations to “pure”
additive utility functions (Yoon and Hwang, 1995).
3 INTRODUCING THE VALUE
OF INFORMATION SECURITY
INVESTMENT AS A CONCEPT
For any organisation, there are a number of
questions relating to Information Security
investments. Organisations are interested in the
amount of investment that should be made, selection
of the measures that should attract investment and
the effectiveness of any potential investments
(Zhang and Li, 2005). All of these questions can be
addressed via the implementation of a decision
support system that utilises multi-attribute utility
theory under uncertainty. Due to the fact that the
“return” of Information Security investments doesn’t
usually come from increased revenues, decreased
costs or other monetary values, this decision support
system avoids the use of the word “return” but rather
determines the value of Information Security
Investments – not just as a monetary value – by
determining the value added to any attributes, such
as process maturity, time saving or peace of mind
for stakeholders. Hence, we introduce an alternative
method for justifying Information security
expenditure in the form of a Value of Information
Security Investment (VISI) Calculation.
3.1 Risk Management Process
ROSI is best incorporated into a full Risk
Management (RM) process, hence, VISI is also best
utilised in a full RM framework. Pontes et al. (2011)
propose a Comprehensive RM framework that
purposely incorporates an ROSI phase. The phases
identified in the RM framework are; plan, risk
appreciation, ROSI, risk treatment, closing. The RM
framework proposed by Pontes et al., 2011 could
therefore be extended to include a VISI calculation
phase in place of ROSI.
3.2 Proposed Architecture
The main underlying concept of the proposed
solution is to produce a decision support model that
is able to evaluate multiple attributes, multiple
alternatives, ROSI calculations and trade-offs under
uncertainty to identify an optimal security solution
for any particular organisation. The purpose of this
approach is to create a standard method for
calculating VISI and making an informed decision
based upon these VISIC results.
Multi-attribute utility theory will initially form
the analytical core of this system. The multi-attribute
utility model will be fed with data that defines the
objectives of the organisation, the attributes to be
considered in the decision process and the
alternatives. If the number of attributes is denoted as
n and the number of alternatives denoted by m, then
the proposed architecture is designed to be able to
consider n attributes across m alternatives. This
approach allows more sophisticated methods of
dealing with the uncertainty present in VISI
calculations. The initial proposed architecture is
shown in Figure 1.
Figure 1: Initial Proposed Architecture.
3.3 Proposed Validation
In order to validate the decision support model that
is being proposed, a series of realistic and
appropriate case studies need to be considered. The
case studies also need to be reliable and able to
replicate real world behaviour and decisions. The
case studies will be hypothetical, although these will
be validated by the use of industry experts. The case
studies will initially start with a simple, small
decision problem to demonstrate that the solution is
viable. The model complexity will then be increased
to demonstrate applicability of the solution to real
world problems that are currently being encountered.
Yin (2009), characterises case study research as an
empirical inquiry that:
DeterminingtheValueofInformationSecurityInvestments-ADecisionSupportSystem
429
Investigates a contemporary phenomenon
within its real life context; when
The boundaries between phenomenon and
context are not clearly evident.
As the VISI calculation is extremely complex,
comprising of alternatives, objectives, attributes and
their interrelationships, a case study design was
warranted. As stated by Yin (2009), case studies
have at least four different applications, perhaps
most importantly, the explanation of causal links in
the VISI calculation, and to provide enlightenment
to situations in which the VISI calculation being
evaluated has no clear, single set of outcomes. The
case study approach also allows for the use of data
from multiple sources.
3.4 Initial Case Study
This hypothetical case study looks at a medium
sized organisation that is concerned about the safe
use of its sensitive data. As the insider threat is still
identified as a serious threat for organisations
(Pricewaterhouse Cooper, 2014), this case study will
consider the threat agent to be an insider with direct
access to the system and the alternatives available to
be preventive measures that will deter or prevent the
insider from extracting sensitive data. The
psychological reasons for insiders acting maliciously
towards a present or previous employer are outside
the scope of this paper.
Three alternatives available to the organisation
are identified. These include off the shelf software to
prevent unauthorised storage devices being
connected to the network, implementation of a
tracking software that track activity and raise alerts
to suspicious activity and finally the use of database
monitoring software to monitor access to certain
restricted data by insiders.
The organisation is concerned with the following
attributes that it will use to evaluate the decision.
These attributes are cost, lifetime and maturity. As
can be seen, these attributes are measured in
different units and are defined in monetary and time
units as well as a scale. These incommensurate
attributes are just one issue that is met when trying
to perform a VISI calculation to justify information
security expenditure.
This case study will use an additive utility
function U. This form of multi-attribute utility
theory includes individual utility functions U
i
for
each attribute x
j
, usually scaled from 0 to 1 that
reflect the decision-makers trade-offs among the
attributes.
U(x
1
,x
2
,x
3
) =
w
1
U
1
(x
1
)+w
2
U
2
(x
2
)+w
3
U
3
(x
3
)
(2)
Where
w
1
+ w
2
+ w
3
= 1 (3)
Weights may be specified directly, or found
using a swing-weight procedure. Individual utility
functions are assessed using the range of attribute
values across the alternatives being considered in the
decision problem.
In this example, the cost has a unit of GBP, the
lifetime is the expected number of years that the
product will last or be useful and the maturity is a
value given by a third party organisation.
Table 1: Decision Attributes and Alternative 1.
Attribute Alternative 1
Cost £3000
Lifetime 3 years
Maturity Low
Table 2: Decision Attributes and Alternative 2.
Attribute Alternative 2
Cost £9000
Lifetime 5 years
Maturity High
Table 3: Decision Attributes and Alternative 3.
Attribute Alternative 3
Cost £6000
Lifetime 5 years
Maturity Medium
The individual utility for each attribute shall be
assessed in the following way:
Table 4: Individual Utilities for each attribute.
Attribute
Cost U(£3000) = 1, U(£9000) =
0, linear
Lifetime U(3 years) = 0, U(5 years)
= 1, linear
Maturity U(Low) = 0, U(Medium)
= 0.6, U(High) = 1
These assessments for each attribute will allow
for the individual utility for each alternative to be
found below in Tables 5-7.
Table 5: Individual Utility for Alternative 1.
Attribute Alternative 1
Cost 1.0
Lifetime 0.0
Maturity 0.0
KMIS2014-InternationalConferenceonKnowledgeManagementandInformationSharing
430
Table 6: Individual Utility for Alternative 2.
Attribute Alternative 2
Cost 0.0
Lifetime 1.0
Maturity 1.0
Table 7: Individual Utility for Alternative 3.
Attribute Alternative 3
Cost 0.5
Lifetime 1.0
Maturity 0.6
The trade-offs assessment is more difficult to
make. One method is to assess weight ratios. In this
example, the decision maker identifies that cost is
twice as important as lifetime. This may be
interpreted to mean that the change in overall
satisfaction corresponding to a change in cost from
£9000 to £3000 is twice the change in overall
satisfaction relating to a change in lifetime from 5
years to 3 years. Similarly, for maturity, the
decision-maker may judge that the £6000 decrease
in cost is 1 and half times as satisfying as a change
from a high to low rating. These ratios are used to
identify the weights for each attribute in Table 8.
Table 8: Weight Assessments.
Weights
Cost 0.462
Lifetime 0.231
Maturity 0.307
These weights are then used to identify the
overall utility value for each alternative. These are
shown in Table 9 below.
Table 9: Overall Utilities.
Alternatives Utility Value
1 0.462
2 0.538
3 0.6462
Hence, the maximum utility is that of alternative
3. Hence, the approach that should be adopted is that
of the database monitoring software. In order to
assess the VISI calculation of this decision, we first
need to identify the cost to the organisation of not
choosing any of these solutions. This is vital to
ensure that the amount of expenditure is below the
potential losses. The organisation suspects that if a
large-scale breach were to occur that the cost would
exceed £20,000, in the lowest case the cost of the
incident is estimated at £10,000. This shows that
even the minimum cost is higher than that of the
expenditure. When evaluating the expenditure over
the lifetime of the product we see that the expected
losses are still much higher than that of the
expenditure required for alternative 3.
This relatively simple, initial case study shows
the viability of using multi-attribute utility theory to
assess information security expenditure decisions.
So far, the decision has been investigated in terms of
just the attributes cost, lifetime and maturity.
However, information security decisions for
expenditure need further justification. We propose
this comes in the form of a VISI calculation. The
purpose of the VISI calculation is to show the value
of the expenditure. The cost to the organisation if a
major incident occurs is £20,000. The organisation
also identifies the probability of a major incident
occurring to be 0.05 and the probability of a minor
incident occurring to be 0.2. In order to identify the
VISI for each alternative, we need to evaluate the
savings and costs in order to identify the best
possible solution. For alternative 1, the savings are
potentially £17,000; alternative 2 provides a
potential saving of £11,000; alternative 3 provides a
potential saving of £14,000. This should also be
included in the utility theory process, however the
uncertainty present in reality of these potential losses
doesn’t allow for a linear utility function. We
identify the decision-maker to be risk averse and
hence the utility function now will be concave
shaped. In the case of a risk seeking decision-maker,
the utility function would be convex shaped.
We define VISI as follows:
VISI = ∅
(4)
Where represents the breach probability, S
represents the potential savings and I represents the
investment cost of the alternative. This cost should
be considered in a singular unit of time, which is
annually, monthly or weekly for example.
Including this VISI value in to the utility theory
process as another attribute further validates the
expenditure being considered. We are trying to
achieve the highest VISI and hence, the highest
value will be assigned the value of 1, and the lowest
the value of 0 for the assessment of each alternative.
The potential savings and cost of the solution
elements require a pre-determined set of values to
include. By providing a set of values to include in
this process, the calculation will be repeatable and
reproducible as well as comparable across
departments and organisations.
The savings and costs to be included are:
Monetary losses due to downtime.
Monetary losses due to loss of business.
DeterminingtheValueofInformationSecurityInvestments-ADecisionSupportSystem
431
Losses in terms of productivity of employees.
Cost of software, hardware, maintenance and
licencing costs.
Training costs.
This list of savings is currently under review and
will be extended and adapted as required.
4 CONCLUSIONS
This paper has demonstrated that the current state of
the art methods in ROSI calculations for information
security possess deficiencies in producing
repeatable, predictable and reproducible results. Due
to the rapid increase in the number of breaches
organisations now encounter, protecting an
organisation is an important aspect. This solution
should not only meet the criteria in terms of
protection of assets and critical systems but should
also be viewed as cost-effective to stakeholders and
executives.
The expansion of ROI calculations to security is
not viable, due to the nature of security solutions
preventing losses and not generating profit, thus a
specific ROI calculation for security needs to be
used. Due to inaccurate data and inconsistent
methods of calculating and reporting losses, many
ROSI calculations are estimations at best, and only
provide insight when the methodologies are kept
constant – by a single organisation for example. The
current best practice methods are simple
calculations, however these calculations are unable
to deal with hard to define and complex metrics.
This paper presents a multi-attribute utility theory
method instead of using ROSI calculations. This
method allows for decision-making under
uncertainty to be carried out in a logical way even
when considering multiple objectives, attributes and
alternatives. This consistent methodology allows for
repeatable and predictable results that are
reproducible when supplied with the organisational
data and method. This proposed, novel solution is
defined as a Value of Information Security
Investment calculation.
This paper presents the initial solution
framework and an initial case study. The idea of the
initial case study is to show the viability, and ease of
the VISI calculation.
Future work with the VISI calculation will
involve refinement of the calculation method and the
application to more complex, real world decision-
making problems that are encountered in large,
security or government organisations. Also, the RM
framework needs to be expanded to include the VISI
calculation to provide organisations with a complete
picture of their information security position.
Sensitivity analysis also needs to be carried out on
VISI calculations, both coarse and fine, for
sensitivity to large and small changes to the ratios
identified, in order to investigate how these changes
affect the identified solution in terms of the overall
utility.
ACKNOWLEDGEMENTS
This work has been partly funded by ESII, contract
number 3000134902/TIN/21.
REFERENCES
Al-Humaigani, M., and D.B. Dunn. 2003. “A Model of
Return on Investment for Information Systems
Security.” In Circuits and Systems, 2003 IEEE 46th
Midwest Symposium on, 1:483 –485 Vol. 1.
doi:10.1109/MWSCAS.2003.1562323.
Arora, Ashish, Steven Frank, and Rahul Telang. 2008.
“Estimating Benefits from Investing in Secure
Software Development.” Build Security In.
Beautement, A., et al., 2009. “Modelling the Human and
Technological Costs and Benefits of USB Memory
Stick Security.” In Managing Information Risk and the
Economics of Security, edited by M. Eric Johnson,
141–63. Springer US.
Belton, Valerie, and Theodor J. Stewart. 2002. Multiple
Criteria Decision Analysis: An Integrated Approach.
Springer.
Beres, Yolanta, David Pym, and Simon Shiu. 2010.
“Decision Support for Systems Security Investment.”
Manuscript, HP Labs.
European Network and Information Security Agency
(ENISA). 2012. “Introduction to Return on Security
Investment: Helping CERTs Assessing the Cost of
(lack Of) Security.”
Fülöp, János. 2005. Introduction to Decision Making
Methods. Laboratory of Operations Research and
Decision Systems. Computer and Automation
Institute, Hungarian Academy of Sciences.
Gordon, Lawrence A., and Martin P. Loeb. 2002. “The
Economics of Information Security Investment.”
ACM Trans. Inf. Syst. Secur. 5 (4): 438–57.
doi:10.1145/581271.581274.
Hausken, Kjell. 2006. “Returns to Information Security
Investment: The Effect of Alternative Information
Security Breach Functions on Optimal Investment and
Sensitivity to Vulnerability.” Information Systems
Frontiers 8 (5): 338–49. doi:10.1007/s10796-006-
9011-6.
KMIS2014-InternationalConferenceonKnowledgeManagementandInformationSharing
432
Holoman, Kathy, and Aaron Kuzmeskus. 2012. The
Evolution of Return on Security Investment (ROSI).
White Paper WP-SEC-SECURITY ROSI-A4.BU.N.
EN.01.2012.0.01.CC. Schneider Electric.
Howard, R.A. 1968. “The Foundations of Decision
Analysis.” IEEE Transactions on Systems Science and
Cybernetics 4 (3): 211 –219. doi:10.1109/TSSC.1968.
300115.
Ioannidis, C. et al., 2009. “Investments and Trade-Offs in
the Economics of Information Security.” In Financial
Cryptography and Data Security, edited by Roger
Dingledine and Philippe Golle, 148–66. Lecture Notes
in Computer Science 5628. Springer Berlin
Heidelberg.
Keeney, Ralph L., and Howard Raïffa. 1976. Decisions
with Multiple Objectives: Preferences and Value
Tradeoffs. Wiley.
Keeney, Ralph L, and Howard Raiffa. 1993. Decisions
with Multiple Objectives: Preferences and Value
Tradeoffs. Cambridge [England]; New York, NY,
USA: Cambridge University Press.
Keeney, Ralph Lyons. 1975. Examining Corporate Policy
Using Multiattribute Utility Analysis. IIASA.
Korostoff, Kathryn. 2003. “The ROI of Network
Security.” Network World. http://www.networkworld.
com/techinsider/2003/0825techinsiderroi.html.
Levy, Jason. 2005. “Multiple Criteria Decision Making
and Decision Support Systems for Flood Risk
Management.” Stochastic Environmental Research
and Risk Assessment 19 (6): 438–47. doi:10.1007/
s00477-005-0009-2.
Løken, Espen. 2007. “Use of Multicriteria Decision
Analysis Methods for Energy Planning Problems.”
Renewable and Sustainable Energy Reviews 11 (7):
1584–95. doi:10.1016/j.rser.2005.11.005.
Pontes, Elvis, Adilson E., Anderson A. A. Silva, and
Sergio T. 2011. “A Comprehensive Risk Management
Framework for Approaching the Return on Security
Investment (ROSI).” In Risk Management in
Environment, Production and Economy, edited by
Matteo Savino. InTech. http://www.intechopen.com/
books/risk-management-in-environment-production-
and-economy/a-comprehensive-risk-management-
framework-for-approaching-the-return-on-security-
investment-rosi-.
Pricewaterhouse Cooper. 2014. “2014 Information
Security Breaches Survey.” PwC.
Raïffa, Howard, and Robert Schlaifer. 1961. Applied
Statistical Decision Theory. Division of Research,
Graduate School of Business Adminitration, Harvard
University.
Rudolph, Manuel, and Reinhard Schwarz. 2012. “A
Critical Survey of Security Indicator Approaches.” In
2012 Seventh International Conference on
Availability, Reliability and Security (ARES), 291 –
300. doi:10.1109/ARES.2012.10.
Shah., N.H., R.M. Gor, and H. Soni. 2007. Operations
Research. PHI Learning Pvt. Ltd.
Sonnenreich, W. 2006. “Return on Security Investment
(ROSI): A Practical Quantitative Model.” Journal of
Research and Practice in Information Technology 38
(1).
Von Neumann, J, and O Morgenstern. 1947. Theory of
Games and Linear Programming. 2nd Edition. New
York: Wiley.
Yin, Robert K. 2009. Case Study Research: Design and
Methods. 4th ed. Applied Social Research Methods, v.
5. Los Angeles, Calif: Sage Publications.
Yoon, K. Paul, and Ching-Lai Hwang. 1995. Multiple
Attribute Decision Making: An Introduction. SAGE.
Zeleny, Milan. 1982. Multiple Criteria Decision Making.
New York [etc.]: McGraw-Hill.
Zhang, Jin-Ping, and Shou-Mei Li. 2005. “Portfolio
Selection With Quadratic Utility Function Under
Fuzzy Environment.” In Proceedings of 2005
International Conference on Machine Learning and
Cybernetics August 18-21, 2005, Ramada Hotel,
Guangzhou, China, 2529–33. Piscataway, NJ: IEEE.
DeterminingtheValueofInformationSecurityInvestments-ADecisionSupportSystem
433