Ontology and Fuzzy Measures based System for
Information Security Risk Assessment
Raihan Muratkhan, Dauren Kabenov and Dina Satybaldina
L. Gumilyov Eurasian National University, 5 Munaitpasov str., 010000 Astana, Kazakhstan
Abstract. Traditionally the information security risk is defined as a
combination of probability of negative event and a potential impact. But risk of
the breach of information security of the modern organization is the
multidimensional complex concept which is including set of interconnected
variables. Often values of risk factors cannot be precisely defined. Therefore
the information security risk assessment may be defined as a fuzzy problem. In
this paper algorithm of the problem decision of alternative's assessment which
has network-like estimation criteria structure is considered. Connections in
criteria structure are formalized by means of fuzzy integral of Sugeno.
Ontology-based information security knowledge domain has net-like structure
incorporating the most relevant information security concepts (assets, threats,
vulnerabilities and controls) and relations among them. Slots describe
properties of concepts and instances. Each property can be set to a specific
fuzzy value. The estimation criteria structure is network-like and is formalized
as the oriented graph with one source and many drains. The alternative's
estimation result is calculated in criterion-source.
1 Introduction and Related Work
Computational Intelligence (CI) is as advanced low level artificial intelligence that
uses computational adaptation to mimic human logic and reasoning. CI can be
realized in a number of different ways [1]. CI core methods such as artificial neural
networks, fuzzy systems, evolutionary computation, artificial immune systems,
swarm intelligence, and soft computing are described in [2]. The authors analyze a
significant number of research works and provide useful insights into how CI might
be used in an intrusion detection system. It was shown that many of these CI
technologies could easily be adapted to be used as mechanisms that can deal with the
numerous securities.
On the other hand recent studies have shown that the lack of information security
knowledge at the management level is one reason for inadequate or non-existing
information security risk management strategies [3].
Authors of [4] identified information security risk management as one of the top
ten grand challenges in information technology security and demanded sound theories
and techniques to support and enhance existing risk management approaches.
Incomplete knowledge about the information security domain in general and the
Muratkhan R., Kabenov D. and Satybaldina D..
Ontology and Fuzzy Measures based System for Information Security Risk Assessment.
DOI: 10.5220/0004097800770084
In Proceedings of the 9th International Workshop on Security in Information Systems (WOSIS-2012), pages 77-84
ISBN: 978-989-8565-15-0
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
current information security status of the organization was one of the main problems
in information security risk management. A creation of ontologies (e.g. General
Privacy Ontology [5], Security Ontology [6], NRL Security Ontology [7]) that define
the conceptual system of the subject area of information security reduced the severity
of the problem.
The essence of estimation problems consists in the following. On the basis of set
of alternative's characteristics values it is necessary to receive a unique estimation of
this alternative according to the criteria system. The criteria system is considered as
the estimation standard (ideal alternative). The estimation is considered as conformity
degree of this alternative to ideal alternative. Solutions of the multy-criteria problems
of assessment and classification with using of the integral of Sugeno or Choquet are
known [8-10]. However hierarchical problems with not crossed branches of hierarchy
are mainly considered in publications. It is essential restriction for many applications.
Therefore a universal algorithm development for solution of the fuzzy problems of
network structure criteria assessment is importance aspect. In this work the
information security risk estimation algorithm on the base of fuzzy measures is
considered. Reuse-used ontology [6], developed by authoritative experts in the field
of ontological modeling is a positive feature of the approach of risk analysis
2 Decision Algorithm of Fuzzy Assessment Problems with
Network-like Criteria Structure
The considered algorithm provides the problem decision of alternative's estimation
which has network-like estimation criteria structure. Criteria can be quantitative and
qualitative. Quantitative criteria correspond to alternative's characteristics which are
measured in quantitative scale. Qualitative criteria correspond to alternative's
characteristics which are measured in qualitative discrete scales. The algorithm
estimates alternative in several criterion contexts. Use of contexts provides the
description of network-like criteria structure. Criterion contexts formalize the
different view-points on estimation value. Connections in criteria structure are
formalized by means of fuzzy measures Sugeno.
Let's consider formal representation of criteria's system with network-like
structure. We are denoting set of estimation criteria:
},1,{
C
i
QicC ==
(1)
The criteria system is formed by means of relations set. Relations can have various
senses which depend on a problem. For example, relations can reflect functional
dependences of criteria or attributive connections. By analogy to the graphs theory,
the criteria system is the acyclic oriented graph with one source cs (the upper-level
criterion) and with many drains (the lower-level criteria)
},1,{
CD
j
QjCcdCD ==
(2)
without dangling tops. The set CD is considered to be as the set of alternative's
characteristics. This set also is universal set in estimation problem.
78
Using fuzzy measures and fuzzy integrals, it is possible to construct the algorithm
for estimation problem decision. This algorithm will consist of following steps.
1) Transformation of alternative's characteristics values to values which are assigned
on discrete set D
i
CD
Transformation is performed for formation of membership h
i
, which is used for
integration in quantitative criterion c
i
. Transformation is performed differently for
qualitative and quantitative characteristics of alternative.
For Qualitative Characteristics of Alternative. The assignment of qualitative
characteristics is performed directly on discrete set. Estimations of alternative's
qualitative characteristics are represented as membership
]1,0[:
CD
j
Dh
(3)
For Quantitative Characteristics of Alternative. For quantitative characteristics this
transformation is performed by means of linguistic variable, for example as shown in
[9].
To each quantitative criterion c
i
the linguistic variable is attributed:
]1,0[],[:;,1),,{(
maxmin
==
ijijij
D
ij
CD
iji
RRtQjtdT
i
(4)
where ],[
maxmin
ijij
RR - is numerical interval which can be various for different pairs
),(
ij
CD
ij
td
.
The linguistic variable T
i
is composed of functions which correspond to values set
elements of drain-criterion. These functions are assigned on numerical intervals. The
membership for integration is determined as the conformity degree of characteristic's
value to linguistic descriptions
CD
ij
d
:
i
D
iijijijiji
QjrtRReh ,1),,min(],max[{
maxmin
===
(5)
2) Consecutive aggregation of alternative's characteristics values in criteria. To each
criterion the set of fuzzy measures according to elements quantity of criterion
determination set is attributed:
i
i
i
i
SD
D
iji
QjM ,1],1,0[2:)({ ==
μ
(6)
This set is named as criterion contexts. For source-criterion the context is the self
criterion. Fuzzy measures are determined on values set of criterion.
The fuzzy integral of membership along fuzzy measure µ
ij
provides calculation of
an alternative's estimation in criterion c
i
for context j:
;,1,)(
i
i
S
D
iij
Qjhse ==
(7)
where membership is composed of alternative's estimations in criteria from D
i
.
Integration is performed consistently in all tops of criteria structure.
79
3 Security Ontology
Ontology is an explicit formal specification of the terms in explicit specications the
domain and relations among them [11]. Ontologies are useful as means to support
sharing and reutilization of knowledge [12]. This reusability approach is based on the
assumption that if a modeling scheme, i.e., ontology, is explicitly specied and
mutually agreed upon by the parties involved, and then it is possible to share, reutilize
and extend knowledge. Many disciplines now develop standardized ontologies that
domain experts can use to share and annotate information in their fields. Problem-
solving methods, domain-independent applications, and software agents use
ontologies and knowledge bases built from ontologies as data [13].
Reusing existing ontologies may be a requirement if our system needs to interact
with other applications that have already committed to particular ontologies or
controlled vocabularies [13]. There are libraries of reusable ontologies on the Web
and in the literature, for example, the Ontolingua ontology library [14], or the DAML
ontology library [15].
An ontology is a formal explicit description of concepts in a domain of discourse
(classes (sometimes called concepts)), properties of each concept describing various
features and attributes of the concept (slots (sometimes called roles or properties)),
and restrictions on slots (facets (sometimes called role restrictions)) [15]. Ontology
together with a set of individual instances of classes constitutes a knowledge base.
Ontology development includes [15]:
defining classes in the ontology,
arranging the classes in a taxonomic (subclass– superclass) hierarchy,
defining slots and describing allowed values for these slots,
filling in the values for slots for instances.
Security Ontology [6] was used in this work. It was proposed based on the security
relationship model described in the National Institute of Standards and Technology
Special Publication 800-12 [16]. Figure 1 shows the high-level concepts and
corresponding relations of ontology.
Slots describe properties of classes and instances. Each property can be set to a
specific value or a formula to calculate this value of the property. All subclasses of a
class inherit the slot of that class. On the other hand, subclasses can have their own
properties.
Fragments of the ontology, including the structure and properties of concepts are
the basis for description of the situation, which is determined by the input data. The
concepts and relationships defined by the input conditions are introduced in addition
to these ontology fragments.
The security ontology comprises about 500 concepts and 600 formal restrictions;
to ensure a minimal encoding bias the ontology is represented by either graphical,
textual, or description logics (DL) representations, which were used to represent
knowledge in a structured, formal, and reasonable form.
80
Fig. 1. High-level concepts and corresponding relations of ontology [6].
4 Proposed Concept of Information Security Risk Assessment
System
Most existing risk analysis models are based on quantitative techniques such as
Monte Carlo Simulation and Annual Loss Expectancy. However, the information that
is related to most uncertainty factors is not numerical. FST provides an approximate
model for the evaluation of the risk faced by EC projects through a linguistic
approach. The procedure for fuzzy risk analysis is based on the works from [17] that
consisted of five steps: risk identification, natural language representation, fuzzy
assessment aggregation, and linguistic approximation.
The first step is to conduct risk identification and compile a list of the most
significant uncertainty factors and their descriptions. Before conducting fuzzy risk
analysis, one must identify the components of risks for the assets on the security
ontology base.
Knowledge base (metaclasses of the assets, threats, vulnerabilities, evaluation
criteria, etc) are developed with based on a comprehensive literature review and
interviewed with practicing auditors of an information security management system
[18,19].
Information security risk is a function of the likelihood of a given threat-source’s
a particular potential vulnerability, and the resulting impact of that adverse event on
the organization. To determine the likelihood of a future adverse event, threats to an
IT system must be analyzed in conjunction with the potential vulnerabilities and the
controls in place for the IT systems. The adverse impact of a security event can be
described in terms of loss or degradation of any, or a combination of any, of the
following three security goals: integrity, availability, and confidentiality. Therefore, it
is necessary to consider a large set of information sources and uncertainty of the
information. On the second step (natural language representation) the linguistic terms
sets, characterizing values of input parameters (asset value, asset criticality, threat
level, vulnerability level, impact level) and an output parameter (risk level) are
defined (see Table 1).
81
Table 1. Membership functions and the triangular fuzzy numbers of the linguistic term.
Likelihood Severity
Unlikely Low =1-4x 0x0.25
Medium Moderate =2(1-2x) 0.25x0.5
Likely High =3-4x 0.5x0.75
Very likely Critical =4x-3 0.75x1
In third stage (fuzzy assessment aggregation) an aggregate of several experts’
fuzzy assessment is performed by using the fuzzy average operation for aggregate
method. Adequacy of expert's preferences formalization depends on mathematical
properties of aggregation operator. Most effective formalization tool for connections
between criteria is the fuzzy measure Sugeno. And most effective aggregation tool is
the fuzzy integral.
From the mathematical view-point the Sugeno and Choquet integrals are in detail
considered in [10]. The Sugeno and Choquet integrals provide various properties of
aggregation procedure which depend from properties of fuzzy measure [20]. For
probability measure (0 = l) the fuzzy integral is equivalent to additive aggregation.
For possibility measure (1 - = l) the fuzzy integral is equivalent to maximum of
membership (fuzzy logic "OR"). For necessity measure (0 >> l) the fuzzy integral is
equivalent to minimum of membership (fuzzy logic "And"). Other values l will
determine other aggregation properties.
As the result of the calculated fuzzy weighted average is a fuzzy number, it is
necessary to translate it back into linguistic terms for easy interpretation. The goal of
linguistic approximation is to find the linguistic term with the closest possible
meaning to that of a defined fuzzy set. There are three techniques in linguistic
approximation: best fit, successive approximation, and piecewise decomposition [17].
The difference between these three techniques was discussed by Schmucker [21]. In
the present study, the best fit method is adopted because it is easy to understand and
easy to implement on computers [21]. This method is based on the ‘‘Euclidean
distance’’ between two fuzzy sets, as proposed in [22]. The model then assigns the
appropriate natural language expression to the lowest Euclidian distance associated
with fuzzy set X.
Fig. 2. Structure of assessment problem decision.
The proposed structure of risk assessment system does not differ from classical
82
structure of expert systems [20]. It is shown in figure 2.
The algorithm is software implemented by authors. The example of the
information security risk assessment of the organization according to international
standards requirements and preferences of the information assets owner is considered.
5 Conclusions and Future Work
This paper presented an ontology-based approach that addresses the problem of
information security risk assessment. The considered algorithm provides the problem
decision of alternative's estimation which has network-like estimation criteria
structure. Quantitative criteria correspond to alternative's characteristics which are
measured in quantitative scale. Qualitative criteria correspond to alternative's
characteristics which are measured in qualitative discrete scales.
The algorithm estimates alternative in several criterion contexts. Use of contexts
provides the description of network-like criteria structure. Connections in criteria
structure are formalized by means of fuzzy measures Sugeno.
The information security risk assessment focused on risk identification, analysis,
and prioritization. Less attention was given to the risk management planning,
resolution, and monitoring. Further research should be conducted into such risk
management planning. In addition, risk monitoring should be conducted regularly to
track the status of the identified risks. With such insight and improvement, the
proposed approach could be further enhanced to handle the functionality of risk
management.
References
1. Mazur, S., Blasch, E., Chen, Y. and Skormin, V.: Mitigating Cloud Computing Security
Risks usinga Self-Monitoring Defensive Scheme. Distribution Statement A: Approved for
Public Release, (2011) 88ABW-2011-3983.
2. Wu, S. X. and Banzhaf, W.: The use of computational intelligence in intrusion detection
systems: A review. Applied Soft Computing, Vol. 10, (2010) 1–35.
3. Straub, D. and Welke, R.: Coping with systems risk: Security planning models for
management decision making. MIS Quarterly, Vol. 22(4), (1998) 441-469.
4. Smith, S. and Spaord, E.: Grand challenges in information security: Process and output.
IEEE Security & Privacy, Vol. 2(1), (2004) 69-71.
5. Hecker, A., Dillon, T., and Elizabeth, C.: Privacy Ontology Support for E-Commerce,
Internet Computing, Issue No. 2, (2008) 54 – 61.
6. Fenz, S. and Ekelhart, A.: Formalizing information security knowledge. ASIACCS ‘09:
Proceedings of the 2009 ACM symposium on Information, computer and communications
security, ACM, (2009) 183-194.
7. Kim, A, Luo, J. and Myong, K.: Security Ontology for Annotating Resources. Naval
Research Lab, NRL Memorandum Report, NRL/MR/5540-05-641: Washington, D.C.,
(2005).
8. Yeong Min Kima and Chee Kyeong Kimb.: Fuzzy based state assessment for reinforced
concrete building structures. Engineering Structures, Vol. 28. 9 (2006) 1286-1297.
83
9. Magyla, T.: The evaluation implementation impact of centralized traffic control systems in
railways. Kaunas University of Technology - Transport, Vol.17, No. 3 (2002) 96-102.
10. Pham, T. and Wagner, M.: Similarity normalization for speaker verification by fuzzy
fusion. Pattern Recognition, vol. 33 (2000) 309-315.
11. Gruber, T. R.: Toward principles for the design of ontologies used for knowledge sharing.
International. J. of Human-Computer Studies, Vol. 43(5-6), (1195) 907–928.
12. Decker, S., Erdmann, M., Fensel, D. and Studer, D. Ontobroker: Ontology based access to
distributed and semi-structured information. DS-8: Semantic Issues in Multimedia Systems,
1999.
13. Noy, Natalya F. and McGuinness, Deborah L.: Ontology Development 101: A Guide to
Creating Your First Ontology'. Stanford Knowledge Systems Laboratory Technical Report
KSL-01-05 and Stanford Medical Informatics Technical Report SMI-2001-0880, March
2001.
14. The Ontolingua ontology library. [Online] Available from:
http://www.ksl.stanford.edu/software/ontolingua/
15. The DAML ontology library. [Online] Available from: http://www.daml.org/ontologies
16. NIST. An Introduction to Computer Security – The NIST Handbook. Technical report,
NIST (National Institute of Standards and Technology) (1995). Special Publication 800-12.
17. Ngai, E. W. T. and Wat, F. K. T.: Fuzzy decision support system for risk analysis in e-
commerce development. Decision Support Systems, Vol. 40 (2005) 235–255.
18. Satybaldina, D.: A Fuzzy Rule Knowledge-Based System for Information Security Risk
Analysis. American Index of Central Asian Scholarship (AICAS), Vol.1, 2 (2010) 61–
67.
19. Satybaldina, D.: A formalized approach to determining security requirements. Proceeding
of the Third Congress of the World Mathematical Society of Turkic Countries. Almaty,
(2009) 215-221.
20. Sveshnikov,S. and Bocharnikov, V.: Contextual algorithm for decision of fuzzy estimation
problems with network-like structure of criteria on the basis of fuzzy measures Sugeno.
MPRA Paper No. 17351, posted 17 (2009). [Online] Available from: http://mpra.ub.uni-
muenchen.de/17351
21. Schmucker, K. J.: Fuzzy Sets, Natural Language Computations and Risk Analysis,
Computer Science Press, Rockville, MD (1984).
22. Dubois, D. and Prade, H.: Fuzzy Sets and Systems, Academic Press, New York (1980).
84