Information Security Governance Analysis Using
Probabilistic Relational Models
Waldo Rocha Flores and Mathias Ekstedt
Royal Institute of Technology, 100 44, Stockholm, Sweden
Abstract. This paper proposes the use of Probabilistic Relational Models
(PRM) for analyzing dependencies between Information Security Governance
(ISG) components and its impact on process capability of mitigating
information security vulnerabilities. Using the PRM enables inference between
different ISG components expressed in probabilities, and also inference on the
process capability. A concrete PRM which exemplifies how to assess the
capability of the access control process is further presented, and thus showing
how the PRM can be adapted to fit the analysis of a specific process in an
organizational environment.
1 Introduction
Information Security Governance (ISG) provides a holistic approach to information
security, and considers management commitment and leadership, organizational
structures, user awareness, policies, processes, and technologies, all working together
to ensure information security of enterprise’s assets is maintained at all times [1]. In
order to increase the understanding of dependencies between different ISG mechan-
isms, the ISG structure can be modeled with architecture metamodels. Architecture
metamodels support managers to effectively plan, design and communicate IT and
business related issues, i.e. they provide decision support for managers [2][3]. Model-
ing ISG as architecture metamodels does not only keep track of important ISG me-
chanisms and their internal relationship, it also provides information about any de-
pendencies in the structure. Therefore, the behavior and effect of changes to an ISG
structure can be predicted, and the structure can be established in an enterprise with-
out using trial and error. Conclusions can therefore been drawn on the consequences
in the enterprise given that one ISG mechanism (e.g. process) is not well implemented
or managed, that a certain organizational or human factor has not been considered, or
that an important role has not been assigned. There are several best-practice guide-
lines targeted to support the assessment of an organization’s current ISG. COBIT [4]
for instance, has been suggested as a framework for ISG. COBIT includes security-
related control objectives as check points to achieve security and a maturity model
that can be used to assess the current state of security.
The ISO 27000 is the root for a number of series of international standards for the
management of Information Security. For instance, ISO 27002 is a control-based
framework using the widely known “Plan-Do-Act-Check” approach to ensure infor-
mation security. Its related document, ISO 27004 includes several measures that can
Rocha Flores W. and Ekstedt M..
Information Security Governance Analysis Using Probabilistic Relational Models.
DOI: 10.5220/0003588001420150
In Proceedings of the 8th International Workshop on Security in Information Systems (WOSIS-2011), pages 142-150
ISBN: 978-989-8425-61-4
Copyright
c
2011 SCITEPRESS (Science and Technology Publications, Lda.)
be used to evaluate the impact of implemented security controls and thus assess the
effectiveness of the information security management in an organization [5].
Current architecture frameworks for Business/IT analysis do not provide a model-
ing language that supports assessments in general and for information security is
particular [7]. There is also little research on the use of modeling languages to support
quantitative assessment for information security. One example is the work presented
in [8] where attack steps were defined as a part of a threat. The presented model was
used for analyzing the effect of technical countermeasures and made it possible to
predict quantified expected loss values from the suggested architecture model. How-
ever, technical countermeasures are not enough own its own. The purpose of the
present paper is to propose a model that can be used for analyzing ISG in an organiza-
tional environment. The presented approach includes organizational and human fac-
tors of information security in an ISG structure, and aid assessments of their impact
on security process capabilities, i.e., the effect of mitigating security vulnerabilities.
To support assessments the model is coupled with an assessment and analysis me-
chanism and extended to a Probabilistic Relational Model (PRM) - specifying a tem-
plate for a probability distribution over the architecture model. The classes in this
PRM are abstract and cannot be directly instantiated into an architecture model. They
can however be made concrete if they are specialized into subclasses according to a
set of constraints. If architecture models are instantiations of such concrete classes,
then the process capability to mitigate security vulnerabilities can be inferred from the
architecture model. By applying an instantiated model on real enterprise ISG envi-
ronments weaknesses and potential risks of security vulnerabilities in an organization
can be highlighted, thus supporting decision making and planning. The remainder of
the paper is structured as follows. In the next section PRM which severs as the as-
sessment mechanism is presented. Section three is the locus of this paper’s contribu-
tion and presents the PRM for ISG analysis, and an instantiation of the PRM to show
how the PRM can be used to support capability assessment of the access control
process. Section four concludes the paper.
2 Probabilistic Relational Models
A Probabilistic Relational Model (PRM) specifies a template for a probability distri-
bution over an architecture model [9]. An architecture metamodel M describes a set of
classes,
,…,
. Each class is associated with a set of descriptive attributes and a set
of reference slots (relationships). The set of descriptive attributes of a class X is de-
noted At(X). Attribute A of class X is denoted X.A and its domain of values is denoted
V(X.A). The set of reference slots of a class X is denoted R(X). We use X.υ to denote
the reference slot of X. A reference slot denotes a function from
to
, and its
inverse

denotes a function from
to
. Thus, the fundamental modeling con-
structs are the same as in general conceptual modeling techniques. An architecture
instantiation I (or an architecture model) specifies the set of objects of each class, the
values for the attributes, and the references of the objects. It specifies a particular set
of process, activities, roles, etc. A PRM template describes the metamodel for the
architecture model, and the probabilistic dependencies between attributes of the archi-
tecture objects. A PRM П, together with an instantiated architecture model of specific
143
objects and relations, defines a probability distribution over the attributes of the
objects. This probability distribution is specified similar to a Bayesian network [10]
which consists of a qualitative dependency structure and associated quantitative
parameters. The probability distribution can be used to infer the values of unknown
attributes, given evidence of the values of a set of known attributes.
The qualitative dependency structure of a PRM is defined by associating attributes
X.A with a set of parents Pa(X.A). Each parent of X.A has the form X.τ.B where B
A(X.τ) and τ is either empty, a single reference slot υ or a sequence of reference slots
,…,
such that for all i, Range[
]= Dom[

]. We call τ a slot chain. Note that
when X.τ.B reference attributes external to the class X, it might be referencing a set of
attributes rather than a single one since there multiple instantiated objects of one class
may exist. In these cases, we let X.A depend probabilistically on an aggregated
property over those attributes constructed using operations such as AND, OR, MEAN
etc. Considering the quantitative part of PRMs, given a set of parents for an attribute,
we can define a local probability model by associating a Conditional Probability Dis-
tribution (CPD) with the attribute, P(X.A|Pa(X.A)). In Fig. 3, fictive numbers are
included to illustrate the CPD table for the instantiated metamodel of the access
control process. Using the fictive numbers gives a value that specifies the probability
that the security process has high capability given that for instance the security culture
in the organization is developed and that the risk awareness is high. We can now
define a PRM Π for a metamodel M as follows. For each class X and each descriptive
attribute A At(X), we have a set of parents Pa(X.A), and a CPD that represents
(X.A|Pa(X.A)). Given a relational skeleton,
(i.e. an instantiated metamodel
without attribute values), a PRM Π specifies a probability distribution over a set of
instantiations I consistent with
:
P
ǀ

Π

.
(1)
where
(X) are the objects of each class in the instantiated metamodel. Hence, the
attribute values can be inferred. A PRM thus constitutes a machinery for calculating
the probabilities of various architecture instantiations. This allows us to infer the
probability that a certain attribute assumes a specific value, given some (possibly
incomplete) evidence of the rest of the architecture instantiation.
3 A Probabilistic Relational Model for Information Security
Governance Analysis
Several definitions to the concept of ISG have been proposed by international associa-
tions, public institutions, and researchers [1][4][5][11][12][13][14]. [4] argues that
ISG is about management commitment and leadership, organizational structures, user
awareness and commitment, policies, processes, and technologies, all working togeth-
er to ensure information security of enterprise’s assets is maintained at all times [1].
[11] and [14] further emphasize assignment of responsibilities and segregation of
duties as important components of ISG. Internal dependencies between ISG compo-
nents such as organizational structure, and security process capabilities has been iden-
144
tified and validated by [15][16]. Adopting these definitions and findings, the proposed
PRM for ISG analysis (Cf. Fig. 1) focuses on how a structure of ISG in an organiza-
tional environment impacts the capability of a process in terms of mitigating vulnera-
bilities. Therefore, a high process capability leads to fewer flaws in an organiza-
tion’s security mechanisms, i.e. vulnerabilities that can be exploited by an attacker.
The qualitative part of the PRM consists of classes, reference slots, attributes and
their parents. A total of six classed were identified Organizational Unit, Process,
Activity, Artifact, Role, and Actor. The main class in the PRM is OrganizationalUnit
that represents an organization. An organization consists of processes and in our case;
processes to mitigate security vulnerabilities. The OrganizationalUnit has therefore
the reference slot ConsistOf whose range is the class Process. Each Process further
consists of a set of activities that defines a process and takes and creates artifacts such
as security policies, back-up storage, etc. This is represented by the two classes
Activity and Artifact with an IsapartOf reference slot for the Activity class and an
ExistsIn reference slot for the Artifact class. In a PRM, classes can further be specia-
lized through inheritance relationships. The classes are related to each other using
subclass relation. For instance, the AccessControlProcess is a subclass of Process
(AccessControlProcess << Process) and then Process class is a superclass of Ac-
cessControlProcess. In the PRM this inheritance relationships is represented by an
IsakindOf reference slot.
A role (e.g. a security manager) is assigned to a process. This relation is illustrated
by the class Role with an IsResponsibleFor/IsAccountableFor reference slot whose
range is the class Process. This Role class has further an IsakindOf reference slot
illustrating that there exist several specializations of a role. The SecurityManageRole
is for instance a subclass of the class Role (SecurityManagerRole << Role), and
Role is then a superclass of SecurityManagerRole. A role is further a resource in the
organization; this relation is represented by the class Role with an IsaResourceIn
reference slot whose range is the OrganizationalUnit. Finally, an actor fills the role,
and is illustrated in the PRM with the class Actor
and a FillsA reference slot with the
range Role.
Regarding the attributes in the PRM, The Process capability attributes is first and
foremost influenced if formal processes are effectively implemented. Further, the
capability of a process is influenced by an organization’s security culture, i.e. shared
attitudes, values, goals, and practices related to information security. The organization
further need to promote and communicate security awareness, establish security
awareness programs, provide education of employees about security policies, etc.
[11][13][14][15][18].
Internal efficiency in terms of the execution of activities, production of artifacts
and the capability of roles has earlier been identified to influence process capability in
[3][19]. We therefore, include an attribute considering if the process is efficiently
managed. The effective implementation of security processes in organizations is
strongly influenced by organizational factors such as top management support,
organizational size, how reliant the organization is on information technology, i.e. IT
reliance, and the environmental uncertainty [15][16][18]. Top management support
may take the form of guidance during planning, participation during design or
involvement during deployment. Besides the ability to secure adequate resources, top
management can also encourage positive user attitude towards the use of information
security. The size of the organization matters as smaller organizations suffer from a
145
lack of human and financial resources. With the lack of funds, smaller organizations
often implement their security processes throughout the organization in a less optimal
way. Depending on the type of industry the organization is active in, the requirements
for security differ. For instance, information security plays a more strategic and criti-
cal role in the financial industry were the environment is highly competitive and vola-
tile [18].
Competence is another critical factor for the effectiveness of implementing
information security [18]. However, [3] argues that competence influences an actor’s
decision making ability, and as competence is defined as experience and how often an
actor works with a subject by [15], these variables are included as attributes in the
Actor class and Role class, which in turn influence if the process is efficiently
managed in the process class.
In Table 1, the attributes for the abstract PRM are shown in this form, i.e. attributes
together with the slot chains defining their parents. In Fig. 1, the complete abstract
PRM is presented including classes, reference slots between classes, the attributes of
each class and the attribute relations. The attributes of the PRM are all defined with
scales. Excerpts of five attribute scales are presented in Fig. 2.
Table 1. Attributes for the abstract PRM shown together with the slot chains defining their
parents.
Attribute
Process.Capability
OrganizationalUnit.SecurityCulture
OrganizationalUnit.SecurityAwareness
Process.IsEffectivelyImplemented
Process.IsEfficientlyManaged
Process.IsEffectivelyImplemented
Process.HasTopManagementSupport
OrganizationalUnit.DegreeOfITReliance
OrganizationalUnit.EnvironmentalUncertainty
OrganizationalUnit.Size
Process.IsEfficientlyManaged
Artifact.Existence
Activity.IsExecuted
Role.HasRoleCapability
Role.HasRoleCapability
Role.IsAssigned
Actor.HasDecisionMakingAbilities
Actor.HasDecisionMakingAbilities
Actor.Competence
ActorIsAwareOfRisks
Actor.Competence
Actor.DegreeOfExperience
Actor.Frequency
Regarding the quantitative part of the PRM, for each attribute local probabilities are
presented using CPDs tables. Fictive numbers are used in Fig. 3 to demonstrate how
146
an instantiated PRM (combined with the proposed dependency model) allows infe-
rence of the process capability.
Fig. 1. The proposed PRM presenting classes, reference slots, attributes and their parents.
Fig. 2. Examples of attribute definitions, and model calculation scales for the two classes
Proce
ss and Organizational Unit.
3.1 Application to the Access Control Process
A simplified demonstration of how a modeler can use the PRM for ISG analysis
applied to the access control process is presented in Fig 3. The access control process
consists of 6 activities that are needed to be executed, 4 artifacts that should exist, and
2 roles that should be assigned in the process. In the evaluation of the fictive medium-
Process
Organizational Unit
Artifact
ExistsIn
SecurityCulture
Exists
Legend
Class
Attribute
ReferenceSlot
[Domain]
[Range]
[Attribute dependency]
[Child]
[Parent]
FillsA
1..*
1
1..*
1..*
IsaResourceIn
1
IsapartOf
ResponsibleFor/
IsAccountableFor
HasTopManagement
Support
1..*
Capability
0..*
0..*
Activity
IsapartOf
1..*
1
Role
Actor
IsAssigned
Competence
Frequency
DegreeOfExperience
IsExecuted
1..*
1
Size
Enviromental
Uncertainty
IsEffectivelyImple
mented
DegreeOfITreliance
HasDecisionMaking
Abilities
IsEfficientlyManaged
ConsistsOf
0..*
0..*
IsakindOf
1..*
1..*
SecurityAwareness
IsakindOf
0..*
0..*
IsAwareOfRisks
HasRoleCapability
Class Attribute Definition Model scale
The degree of Top management support for
information security.
Subjectively evaluated using perceptual responses
If more than 50% of the a c tivities relate d to a
process is executed
{True, False}
If more than 50% of the artifacts related to a
process exists
{True, False}
If the role is capable of managing the process or
is accountable for the process
{True, False}
The degree of security culture in the organization
Subjectively evaluated using perceptual responses
The de
g
ree of securit
y
awareness in the
Subjectively evaluated using perceptual responses
Is Efficiently
managed
Process Has Top
Management
Support
{True, False}
Security
Awareness
{Low, Medium, High}
Organisational
Unit
Security
Culture
{Not developed, Developed,
Highly developed}
147
Fig. 3. The instantiated PRM for access control with CPD tables and fictive numbers.
sized company AB Nordic Energy, the modeler found that 4 out of 6 activities were
executed, and 2 out of 4 artifacts existed in the process. Thus, 67% of the suggested
process activities were executed and 50 % of the artifacts existed. As both values are
more than 50%, these two values yield a True state for activities and a False for
artifacts. Further, at least one of the business manager (Mrs. Adams) and security
..
.. .. ..
..
T F T F .. .. .. T F T F
T 10.90.90.8..............
F 00.10.10.2..............
SmallSize Large
LDegreeOfITReliance H
IsEffectivelyImplemented
TF
HasTopManagementSupport
EnvironmentalUncertainty T F
.. Not Developed
.. .. ..
.. ..
T F T F .. .. .. T F T F
H 10.90.90.7.. ...... .. ....
M 00.10.10.2.. ...... .. ....
L 0000.1.. ...... .. ....
F
H
Highly developedSecurityCulture
L
TF
SecurityAwareness
Capability
IsEffectivelyImplemented
IsEfficientlyManaged
T
..
TF..TF
T 0.9 0.8 .. 0.2 0.1
F 0.1 0.2 .. 0.8 0.9
L
IsAwareOfRisks
HasDecision
MakingAbility
Competence H
High Low
Frequently .. Unfrequently
H0.9 .. 0
M0.1 .. 0.1
L0 .. 0.9
Competence
IsExperienced
Frequency
IsAssigned T F
T0.90.6
F0.10.4
HasRoleCapability
HasDecisionMakingAbilities
T
..
TF.. T F
T10.7..0.20
F00.3..0.81
HasRoleCapability
ActivityIsExecuted T F
FT
ArtifactExistence
IsEfficiently
Managed
148
manager (Mr. Brown) is responsible for managing the process. The decision making
ability (based on their competence and awareness of risks) of the two actors was
assessed to False and True. Therefore, by combining the fictive probabilistic values
in the local CPD table, the attribute IsEfficientlyManaged yielded a True state.
As AB Nordic Energy is a medium-sized company active in an uncertain
environment and at the same time is Medium reliant on IT, but have top management
support for security endeavors, the attribute IsEffectivelyImplemented yielded a True
value based on the local CPD table for the attribute. Finally, by collecting data for the
last two organizational factors (security culture and security awareness) and combin-
ing these with the already collected data in the CPD table for the attribute process
capability, a state of High was yielded.
This simplified example of how the PRM can be put into practice has shown that
by collecting data from an enterprise ISG structure, using constraint such as those
presented in Fig 2., a value of process capability can be inferred. Further, different
components in the ISG PRM structure can be analyzed and their dependency can be
calculated.
4 Concluding Remarks and Future Work
A literature review was conducted to extract important variables for the governance of
information security, their relation to organizational and human factors and impact on
security process capability. The main contribution with this paper is that we have
argued and demonstrated the general feasibility of performing ISG analysis using
PRMs. We have also provided an instantiated PRM of the access control process and
thus shown how this analysis can be adapted to fit the analysis of a specific process in
an organizational environment. Of interest for future work, and parts of ongoing re-
search, is the enhancement of the PRM; focusing on more specialization of security
processes to provide a PRM that can be used to analysis and calculate the aggregated
capability value of several processes combined such as the risk management process,
patch management process, incidents management process etc. The greatest part of
future research is however to validate the qualitative structure and the attribute rela-
tions in the PRM so that the CPDs can be set with real empirical data. The data will
be collected from case studies combined with surveys with the aim to improve the
accuracy of the predictive function that the PRM can provide.
References
1. Von Solms, S H. Information Security Governance - Compliance Management vs. Opera-
tional Management. Computer & Security. September 2005, pp. 443-447.
2. Winter, R and Fischer, R. Essential layers, artifacts, and dependencies of enterprise archi-
tecture. Journal of Enterprise Architecture, volume 3. 2007, pp. 7-18.
3. Lagerström, Robert, et al., et al. A Method for Creating Enterprise Architecture Metamo-
dels - Applied to System Modifiabiltiy Analysis. International Journal of Computer
Science and Applications, Vol. 6, No. 5. 2009, pp. 89-120.
4. ISACA. Control Objectives for Information and related Technology: ISACA, 2007.
149
5. ISO/IEC. ISO/IEC 27002:2005 Information technology - Security techniques - Code of
practice for information security management. Switzerland : ISO/IEC, 2005.
6. ISO/IEC, JTC1/SC27. Common criteria for information technology security evaluation –
part 1: introduction and general model: ISO/IEC, 2006.
7. A Modeling Language for Interoperability Assessments. Ullberg, J, Johnson, P and
Buschle, M. Stockholm : Springer, 2011. IWEI, Lecture Notes in Business Information
Processing. pp. 61-74.
8. Sommestad, Teodor, Ekstedt, Mathias and Johnson, Pontus. A probabilistic relational
model for security risk analysis. Computers & Security 29. 2010, pp. 659-679.
9. Learning probabilistic relational models. Friedman, N, et al., et al. 1999. Proceeding of the
16th International Joint Conference on Artificial Interlligence. pp. 1300-1309.
10. Jensen, Finn V. An introduction to Bayesian networks. New York : Springer-Verlag, 1996.
11. ISACA. Information Security Governance: Guidance for Board of directors and Executive
management 2nd Edition: ISACA, 2006.
12. ISF. The standard of Good Practice for Information Security: Information Security Forum,
2007.
13. NIST. Special Publication 800-100 Information Security Handbook: A Guide for Manag-
ers: NIST, 2006.
14. US-CERT. Governing for Enterprise Security (GES) Implementation Guide: US-CERT,
2007.
15. Chang, Shuchih E and Ho, Chienta B. Organizational factors to the effectiveness of im-
plementing information security management. Industrial Management and Data Systems,
Vol. 3 No 3. 2006, pp. 345-361.
16. Kankanhalli, Atreyi, et al., et al. An integrative study of information systems security
effective-ness. Journal of International Journal of Information Management 23. 2003, pp.
139-154.
17. Knapp, Kenneth, et al., et al. Information Security Effectiveness: Conceptualization and
Validation of a Theory. International Journal of Information Security and Privacy, Volume
1 Issue 2. 2007, pp. 37-60.
18. Chang, Shuchih and Lin, Chin-Shien. Exploring organizational culture for information
security management. Industrial Management and Data Systems, Vol. 107 No 3. 2007, pp.
438-458.
19. Simonsson, M, Johnson, P and Ekstedt, M. The effect of IT Governance Maturity on IT
Governance Performance. Information Systems Management. December 2010, pp. 10-24.
150