matching cyber security ontologies through genetic

Research ArticleMatching Cyber Security Ontologies through GeneticAlgorithm-Based Ontology Alignment Technique

Weiwei Lin 12 and Reiko Haga 3

1School of Big Data and Artificial Intelligence Fujian Polytechnic Normal University Fuqing 350300 China2Engineering Research Center for ICH Digitalization and Multi-source Information Fusion Fujian Province UniversityFuqing 350300 China3CommScope Japan KK Nagatacho Tokyo 100-0014 Japan

Correspondence should be addressed to Weiwei Lin linww_cnhotmailcom

Received 23 September 2021 Revised 23 October 2021 Accepted 27 October 2021 Published 30 November 2021

Academic Editor Pei-Wei Tsai

Copyright copy 2021Weiwei Lin and ReikoHaga-is is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

Security ontology can be used to build a shared knowledge model for an application domain to overcome the data heterogeneityissue but it suffers from its own heterogeneity issue Finding identical entities in two ontologies ie ontology alignment is asolution It is important to select an effective similarity measure (SM) to distinguish heterogeneous entities However due to thecomplex semantic relationships among concepts no SM is ensured to be effective in all alignment tasks-e aggregation of SMs sothat their advantages and disadvantages complement each other directly affects the quality of alignments In this work we formallydefine this problem discuss its challenges and present a problem-specific genetic algorithm (GA) to effectively address it Weexperimentally test our approach on bibliographic tracks provided by OAEI and five pairs of security ontologies -e results showthat GA can effectively address different heterogeneous ontology-alignment tasks and determine high-quality securityontology alignments

1 Introduction

Security ontology builds a shared knowledge model for aninformation systemrsquos security area to facilitate the estab-lishment of trust relationships [1] Figure 1 shows an ex-ample of security ontology An oval denotes a concept suchas SecurityProtocol or ProtocolEncryption -e arrow be-tween two concepts denotes a subsumptive relationship forexample ProtocolSignature is subsumed by Secur-ityProtocol A concept might have properties such as theXACML and ACL properties of ProtocolAccessControlHowever security ontologies have different applicationrequirements and bias interest which causes the ontologiesthemselves to suffer from the heterogeneity problemFinding identical entities in two security ontologies iesecurity ontology alignment is a solution to this issue [2 3]It is important to use a similarity measure (SM) to distin-guish heterogeneous entities when aligning security

ontologies However due to the complex semantic rela-tionships among concepts no SM is effective in all contextsHence it is important to aggregate SMs so that their ad-vantages and disadvantages complement each other

-e most flexible way to aggregate SMs is the parallelframework which assigns a weight for each SM to obtain thefinal alignment During this procedure each SMrsquos similaritymatrix is calculated whose rows and columns respectivelyrepresent two ontologiesrsquo entities and whose elements aretheir similarity values -e aggregated matrix is determinedby aggregating all the matrices with the weighted meanstrategy A threshold is used to filter elements with lowsimilarity values to obtain the final matrix which is decodedto the ontology alignment It is a complex problem to de-termine the optimal aggregating weight set for SMs sincethere are many local optimal solutions Genetic algorithm(GA) [4 5] is a classic global optimization algorithm whichis adept at solving the optimization problem without the

HindawiSecurity and Communication NetworksVolume 2021 Article ID 4856265 7 pageshttpsdoiorg10115520214856265

information of the objectiversquos gradient Being inspired by itssuccess in the complex optimization domains [6 7] we builda mathematical model under a parallel aggregating frame-work to define the security ontology alignment problempropose a problem-specific GA to address it and determinehigh-quality security ontology alignments

-e remainder of this paper is arranged as followsSection ldquoPreliminariesrdquo defines the security ontologyalignment and similarity measure Section ldquoGenetic Algo-rithm to Integrate Security Ontologiesrdquo describes the GA-based alignment technique Experimental results are dis-cussed in section ldquoExperimentrdquo and section ldquoConclusionrdquorelates our conclusions

2 Preliminaries

21 Security Ontology Alignment Security ontology consistsof concepts properties and axioms and an ontologyalignment is a mapping set A mapping is a 3-tuple (c1 c2sim) where c1 and c2 are two ontologiesrsquo entities and sim istheir similarity [8 9] Aligning ontologies require us to findthe correspondence between two ontology entities to bridgetheir semantic gap As shown in Figure 2 the input ofontology alignment is a pair of ontologies After usingdifferent SMs to determine the corresponding similaritymatrices GA is used to optimize their aggregating weights toobtain the final alignment

A security ontology alignmentrsquos quality can be measuredwith metrics in the information retrieval domain [10]

recall |RcapRA|

|R|

precision |RcapRA|

|A|

f minus measure 2 precision middot recallrecall + precision

(1)

where A and RA are respectively an alignment and ref-erence alignment and denotes a setrsquos cardinalities Here f-measure is the harmony mean of recall and precision Onthis basis the security ontology alignment problem has theobjective to maximize the f-measure and the decisionvariable is X (x1 x2 )T where xi isin [0 1] i 1 2 is the ith SMrsquos aggregating weight and 1113936 xi 1 In thiswork we choose the weighted average strategy to aggregatethe SMs which is the most popular and flexible method inthe domain of information fusion of combining SMs -eother aggregating mechanisms such as those in the field ofevidential reasoning and fuzzy reasoning could be alsoapplied which is one of our future works

22 Similarity Measure SM can generally be categorized aseither syntactic linguistic or taxonomy SM [11 12] whichwe describe as follows

Syntactic SM calculates the similarity of two stringsthrough their edit distance We use the Levenshtein distance[13]

Levenshtein s1 s2( 1113857 max0 min s1

11138681113868111386811138681113868111386811138681113868 s2

111386811138681113868111386811138681113868111386811138681113872 1113873 minus d s1 s2( 1113857

min s11113868111386811138681113868

1113868111386811138681113868 s21113868111386811138681113868

11138681113868111386811138681113872 1113873

(2)

where |s1| and |s2| are the respective character numbers ofstrings s1 and s2 and d(s1 s2) is their edit distance

Linguistic SM utilizes an electronic dictionary to mea-sure the similarity of two words We useWordNet [14 15] asthe electronic knowledge base Linguistic similarity is de-fined as

Linguistic w1 w2( 1113857 maxc1isinsen w1( )c2isinsen w2( )

sim c1 c2( 11138571113858 1113859

(3)

where w1 and w2 are words derived from two entities andsen(wi) denotes the number of meanings of wi

SecurityAssertion

Security Protocol SecurityEncryption SecuritySignature Security Algorithm

ProtocolSignature

OpenPGP-SignatureXML-Signature

ProtocolAuthentication

KerberosUsername

ProtocolCanonization

Canonization

ProtocolEncryption

OpenPGP-EncryptionXML-Signature

ProtocolAccessControl

XACMLACL

Component

Elements

Data

AlgAsymmetric

RSADSA

AlgSymmetric

DESAES

AlgHash

SHA1MD5

AlgCanonization

XMLC14N

SecurityMichanism

DigitalDigestDigitalSignature

SecurityToken

UsernameTokenX509CertificateKerberosTickethellip

hellip

hellip

hellip hellip hellip hellip hellip

hellip

hellip

hellip

Figure 1 An example of security ontology

2 Security and Communication Networks

Taxonomy SM uses the context of concepts c1 and c2 todetermine their similarity [16 17]

Taxonomy c1 c2( 1113857 Levenshtein super1 super2( 1113857 + avg Levenshtein subi subj1113872 11138731113966 1113967

2 (4)

where super1 and super2 are the superclasses of c1 and c2respectively and subi and subj are respectively their i thand j th subclasses In particular the taxonomy SM de-termines the similarity value by calculating the averagesimilarity of two conceptsrsquo parent pair and all their directsubclass pairs

3 Genetic Algorithm to IntegrateSecurity Ontologies

31 Encoding Mechanism In this work we use binarycoding [18] to reduce the evolutionary operationrsquos com-putational complexity Considering that the coding

Ontology Alignment

OOprime

e1

e1

e2

e2

ei

ej en

em

OOprime

e1

e1

e2

e2

ei

ej en ej en

em

OOprime

e1

e1

e2

e2

ei

em

02

01

01

02

01 01

03

03

01

02 01

03

02

03

06

Genetic Algorithm based Aggregation

Similarity measure2 hellip

hellip

Similarity measurenSimilarity measure1

Similarity Matrix2 Similarity MatrixnSimilarity Matrix1

Source Ontology Target Ontology

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

helliphellip

hellip

hellip

Figure 2 Framework of ontology alignment

Security and Communication Networks 3

information must contain the weight set of SMs we storethem in disguised form by storing the cutting points in thecoding information We sort a set of cutting points Cprime (c1prime c2prime cn

prime) in the ascending order as C (c1 c2 cn)and then we can get the corresponding weight set

wk

c1 k 1

ckminus1 minus ck 1lt klt n + 1

1 minus cn k n + 1

⎧⎪⎪⎨

⎪⎪⎩(5)

-rough calculation we can use n cutting points toobtain n + 1 aggregating weights -is work selects threeSMs so we need to encode the information of two cuttingpoints We use 10 gene bits to represent a cutting pointhence the length of a chromosome is 20 gene bits Figure 3shows an example of the encoding mechanism where twocutting points represent the aggregating weights of the SMand five gene bits are used to encode each cutting point Asshown in the figure a chromosome is decoded to decimal toobtain the cutting point set Cprime which is sorted to obtain thecutting point set C -en weights w1 w2 and w3 are cal-culated according to formula (5)

32 Selection -e selection operator is the kernel compo-nent of GA which decides whether a solutionrsquos gene in-formation can persist A solution with a higher fitness valueshould have a greater probability of selection but one with alower fitness value should also have a certain opportunity-is work empirically chooses the classic roulette selectionoperator -e probability of selecting an individual is theratio of its fitness value to the sum of the fitness values of allsolutions hence each individual has the opportunity to beselected If the i th solution has fitness value fi its selectionprobability is fi1113936 fi

33 Crossover -e crossover operator mixes the genes oftwo parent solutions according to a crossover probabilityWe randomly select a cutting point using the single-pointcrossover operator [19] and two children are generated byswapping the right parts of two parentsrsquo genes

34 Mutation -e mutation operator aims to maintainpopulation diversity which is critical to the algorithmrsquossearching ability -is work selects the locus mutation op-erator [20] which judges whether a gene value should beflipped by generating a random number in [0 1] andcomparing it with the mutation probability

35 Pseudocode of Genetic Algorithm Given the maximumgeneration maxGeneration we present the GA pseudocode

lowastlowast lowast lowast lowast lowast lowast Initializationlowastlowast lowast lowast lowast lowast lowast lowast lowast lowastfori 0 ilt populationlength i + +do

forj 0 jlt populationlength j + +dogeneij random 0 1

end for

end forlowastlowast lowast lowast lowast lowast lowast Evaluationlowastlowast lowast lowast lowast lowast lowast lowast lowast lowastfori 0 ilt populationlength i + +do

evaluation()end forlowastlowast lowast lowast lowast lowast lowast Evolutionlowastlowast lowast lowast lowast lowast lowast lowast lowast lowastgeneration 0whilegenerationltmaxGenerationdo

crossover()mutation()fori 0 ilt populationlength i + +do

evaluation()end forselection()saveElite()generation generation + 1

end while

-e gene values of each individual are initialized as 1 or0 and then the populationrsquos solutions are evaluated In eachgeneration the crossover and mutation operators are suc-cessively applied and all solutions are re-evaluated -eselection operator is then used to determine the populationof the next generation Finally the worst solution is replacedby the best one in the history (ie the elite solution)

36 Experiment We utilized the Bibliographic track fromOAEI (httpoaeiontologymatchingorg) to test the per-formance of our proposal In particular 1XX and 2XX arethe respective testing cases with IDs beginning with 1 and 2In 1XX two ontologies under alignment are exactly thesame except for different OWL restrictions while in 2XXthey are heterogeneous in terms of the entity name andorthe conceptrsquos hierarchical structure We also chose fourpairs of specialized security ontologies for testing (1)Network Security OntologiesmdashNetwork Attack Ontology(NAO) [21] and Ontology-Based Attack Model (NAM)

1Chromosome

051 019

0 0 0 0 0 0 01 1

Cprime

019 051C

019 051

019 032 049AggregatingWeights 0

c1

c2prime

c2

w1 w2 w3

c1prime

c2c1

Figure 3 Example of encoding mechanism


[22] (2) Security Requirement-Related Ontolo-giesmdashSecurity and Domain Ontology for Security Re-quirement Analysis (SDOSRA) [23] and ExtendedOntology for Security Requirements (EOSR) [24] (3)Miscellaneous Security OntologiesmdashOntological approachtoward Cyber Security in Cloud Computing (OCSCC) [25]and Ontology in Cloud Computing (OCC) [26] (4) Ap-plication-Based Security OntologiesmdashSecurity Ontologyfor Mobile Applications (SOMA) [27] and Security On-tology for Mobile Agents Protection (SOMAP) [28] and

Cloud Security Policy (CSP) [29] and Cloud Ontology(CO) [30] -e threshold for filtering the final alignmentwas set as 085 and the configuration of GA was empiricallyset to a maximum 3000 generations crossover rate 06 andmutation rate 002 In the experiment we compared ourapproach with OAEIrsquos participants Table 1 compares theresults in terms of recall and precision and Figure 4compares the f-measures Table 2 shows the results ofusing GA to align the security ontologies -e results of ourapproach were the mean values of 30 independent runs

Table 1 Comparison on OAEIrsquos bibliographic track in terms of recall and precision

Matching system1XX 2XX

Precision Recall Precision RecallEdna 064 100 062 084LogMap 094 096 090 081LogMapLt 056 099 053 083LogMapBio 050 056 052 065GMap 097 100 088 085LogMap-C 058 096 057 081Mamba 090 084 079 076AOT 2014 097 097 093 083OReasoner 087 100 074 084CIDER-CL 100 100 078 091HerTUDA 089 100 090 085MapSSS 089 034 087 027RIMOM 2013 084 100 063 088ServOMap 095 100 067 056StringsAuto 089 034 087 027Synthesis 094 100 081 086XMapGen 084 100 067 078XMapSig 084 100 070 084ASE 058 100 061 085GOMMA 084 100 070 087MEDLEY 072 100 068 084Optima 100 100 085 083ServOMap 100 098 091 076ServOMaplt 100 028 100 045WMatch 084 100 073 085GA 100 100 095 085

00102030405060708091

1XX2XX

GA

CroM

atch

LogM

ap

XMap

CroM

atcher

LogM

ap-C

AOT_

2014

MaasM

tch

ORe

ason

er

HerTU

DA

IAMA

MapSSS

RIMIO

M2013

Strin

gsAu

to

WeSeE

XMapGen

YAM++ ASE

GOMMA

Optim

a

ServOMaplt

Wmath

Figure 4 Comparison of OAEIrsquos Bibliographic track in terms of f-measure


As shown in Table 1 the recall and precision of ourapproach were generally higher than those of OAEI -is isbecause GA is able to effectively jump out of lots of localoptimas and find the optimal aggregating weights fromlarge-scale feasible solutions In particular the precision ofour approach was high which shows that aggregating dif-ferent similarity measures can effectively distinguish het-erogeneous entities

As can be seen from Figure 4 the results of our approachwere the best on 1XX testing cases which shows that GA caneffectively align two ontologies with the same entities andstructures In addition with respect to different heterogeneoustasks on 2XX testing cases our approach was also effectivewhich shows that our approach is able to address the matchingproblem with different heterogeneity characteristics

Table 2 depicts the results of approaches to aligning fivepairs of real security ontologies which show our approachcan achieve a high capacity on all testing cases in terms of thef-measure To sum up our approach was robust ataddressing different alignment tasks and could determinehigh-quality security ontology alignments

4 Conclusions

To ensure communication and cooperation among differentsecurity applications built on security ontologies we proposed aGA-based ontology alignment technique to address the securityontology heterogeneity problem We defined the problemdiscussed its challenges and presented a problem-specificGA toeffectively address it Bibliographic tracks provided by OAEIand five pairs of security ontologies were used to test ourapproachrsquos performance -e experimental results show thatour approach is able to align different heterogeneous ontologiesand determine high-quality security ontology alignments

In the future we are interested in adaptive similarityselection which determines effective and nonconflictingsimilarity measures according to the heterogeneous featuresof two ontologies under alignment Moreover when thenumber of similarity measures is large some strategies toimprove efficiency should be introduced to improve GArsquosperformance

Data Availability

-e data used to support this study can be found in thecorresponding footnotes

Conflicts of Interest

-e authors declare that they have no conflicts of interest inthe work

Acknowledgments

-e authors thank LetPub (httpswwwletpubcom) for itslinguistic assistance during the preparation of this manu-script -is work was supported by the Natural ScienceFoundation of Fujian Province China (grant no2019J01889) the ldquoTiancheng Huizhirdquo Innovation and Ed-ucation Promotion Fund China (grant no 2018A02005)and the Education-Scientific Research Project for Middle-Aged and Young of Fujian Province China (grant noJT180626)

References

[1] S Hacini and R Lekhchine ldquoSecurity ontology for mobileagents protectionrdquo International Journal of Computer 7eoryand Engineering vol 4 no 3 pp 426ndash428 2012

[2] P Shvaiko and J Euzenat ldquoOntologymatching state of the artand future challengesrdquo IEEE Transactions on Knowledge andData Engineering vol 25 no 1 pp 158ndash176 2013

[3] I Osman S Ben Yahia and G Diallo ldquoOntology integrationapproaches and challenging issuesrdquo Information Fusionvol 71 pp 38ndash63 2021

[4] D Whitley ldquoA genetic algorithm tutorialrdquo Statistics andComputing vol 4 no 2 pp 65ndash85 1994

[5] S Katoch S S Chauhan and V Kumar ldquoA review on geneticalgorithm past present and futurerdquo Multimedia Tools andApplications vol 80 no 5 pp 8091ndash8126 2021

[6] X Xue and Y Wang ldquoOptimizing ontology alignmentsthrough a memetic algorithm using both MatchFmeasure andunanimous improvement ratiordquo Artificial Intelligencevol 223 pp 65ndash81 2015

[7] C Kim R Batra L Chen H Tran and R RamprasadldquoPolymer design using genetic algorithm and machine learn-ingrdquo Computational Materials Science vol 186 pp 1ndash6 2021

[8] G Acampora V Loia and A Vitiello ldquoEnhancing ontologyalignment through a memetic aggregation of similaritymeasuresrdquo Information Sciences vol 250 pp 1ndash20 2013

[9] X Xue and Y Wang ldquoUsing memetic algorithm for instancecoreference resolutionrdquo IEEE Transactions on Knowledge andData Engineering vol 28 no 2 pp 580ndash591 2016

[10] C J Van Rijsberge Information Retrieval University ofGlasgow London UK 1975

[11] S Mani and S Annadurai ldquoExplicit link discovery schemeoptimized with ontology mapping using improved machinelearning approachrdquo Studies in Informatics and Controlvol 30 no 1 pp 67ndash75 2021

[12] X Xue and J Chen ldquoMatching biomedical ontologies throughcompact differential evolution algorithm with compactadaption schemes on control parametersrdquo Neurocomputingvol 458 pp 526ndash534 2021

[13] V I Levenshtein ldquoBinary codes capable of correcting dele-tions insertions and reversalsrdquo Soviet Physics Dokladyvol 10 no 8 pp 707ndash710 1966

Table 2 Experimental results on security ontology alignment

Category Testing case Recall Precision f-measureNetwork security ontologies NAO-NAM 082 075 078Security requirement-related ontologies SDOSRA-EOSR 076 090 082Miscellaneous security ontologies OCSCC-OCC 088 095 091

Application-based security ontologies SOMA-SOMAP 084 088 085CSP-CO 086 082 083


[14] G A Miller ldquoWordNetrdquo Communications of the ACMvol 38 no 11 pp 39ndash41 1995

[15] E Geller M Gajek A Reibach and Z Łapa ldquoApplicability ofwordnet architecture in lexical borrowing studiesrdquo Interna-tional Journal of Lexicography vol 34 no 1 pp 92ndash111 2021

[16] M AlMousa R Benlamri and R Khoury ldquoExploiting non-taxonomic relations for measuring semantic similarity andrelatedness in WordNetrdquo Knowledge-Based Systems vol 212pp 1ndash27 2021

[17] X Xue and J Zhang ldquoMatching large-scale biomedical on-tologies with central concept based partitioning algorithm andadaptive compact evolutionary algorithmrdquo Applied SoftComputing vol 106 pp 1ndash11 2021

[18] Y Xue H Zhu J Liang J and A stowik ldquoAdaptive crossoveroperator based multi-objective binary genetic algorithm forfeature selection in classificationrdquo Knowledge-Based Systemsvol 227 pp 1ndash17 2021

[19] F A Zainuddin and M F Abd Samad ldquoComparison ofcrossover in genetic algorithm for discrete-time systemidentificationrdquo International Review of Mechanical Engi-neering (IREME) vol 15 no 2 pp 59ndash66 2021

[20] J Al-Afandi and A Horvath ldquoAdaptive gene level mutationrdquoAlgorithms vol 14 no 1 pp 1ndash18 2021

[21] R P Van Heerden B Irwin and I Burke ldquoClassifyingnetwork attack scenarios using an ontologyrdquo in Proceedings ofthe 7th International Conference on Information-Warfare ampSecurity pp 311ndash324 Seattle WA USA March 2012

[22] J-b Gao B-w Zhang X-h Chen and Z Luo ldquoOntology-based model of network and computer attacks for securityassessmentrdquo Journal of Shanghai Jiaotong University vol 18no 5 pp 554ndash562 2013

[23] A Souag C Salinesi I Wattiau and H Mouratidis ldquoUsingsecurity and domain ontologies for security requirementsanalysisrdquo in Proceedings of the 2013 IEEE 37th AnnualComputer Software and Applications Conference Workshopspp 101ndash107 Washington DC USA July 2013

[24] F Massacci J Mylopoulos F Paci T T Tun and Y Yu ldquoAnextended ontology for security requirementsrdquo in Proceedingsof the International Conference on Advanced InformationSystems Engineering pp 622ndash636 London UK June 2011

[25] T Takahashi Y Kadobayashi and H Fujiwara ldquoOntologicalapproach toward cybersecurity in cloud computingrdquo inProceedings of the 3rd International Conference on Security ofInformation and Networks pp 100ndash109 New York NY USASeptember 2010

[26] L Youseff M Butrico and D Da Silva ldquoToward a unifiedontology of cloud computingrdquo in Proceedings of the GridComputing Environments Workshop pp 1ndash10 Austin TXUSA November 2008

[27] S Beji and N El Kadhi ldquoSecurity ontology proposal formobile applicationsrdquo in Proceedings of the 2009 Tenth In-ternational Conference on Mobile Data Management SystemsServices and Middleware pp 580ndash587 Washington DCUSA May 2009

[28] H Razouki ldquoSecurity policy modelling in the mobile agentsystemrdquo International Journal of Computer Network andInformation Security vol 11 no 10 pp 26ndash36 2019

[29] C Choi J Choi and P Kim ldquoOntology-based access controlmodel for security policy reasoning in cloud computingrdquo7eJournal of Supercomputing vol 67 no 3 pp 711ndash722 2014

[30] K Arbanas and M Cubrilo ldquoOntology in information se-curityrdquo Journal of Information and Organizational Sciencesvol 39 no 2 pp 107ndash136 2015


information of the objectiversquos gradient Being inspired by itssuccess in the complex optimization domains [6 7] we builda mathematical model under a parallel aggregating frame-work to define the security ontology alignment problempropose a problem-specific GA to address it and determinehigh-quality security ontology alignments

-e remainder of this paper is arranged as followsSection ldquoPreliminariesrdquo defines the security ontologyalignment and similarity measure Section ldquoGenetic Algo-rithm to Integrate Security Ontologiesrdquo describes the GA-based alignment technique Experimental results are dis-cussed in section ldquoExperimentrdquo and section ldquoConclusionrdquorelates our conclusions

2 Preliminaries

21 Security Ontology Alignment Security ontology consistsof concepts properties and axioms and an ontologyalignment is a mapping set A mapping is a 3-tuple (c1 c2sim) where c1 and c2 are two ontologiesrsquo entities and sim istheir similarity [8 9] Aligning ontologies require us to findthe correspondence between two ontology entities to bridgetheir semantic gap As shown in Figure 2 the input ofontology alignment is a pair of ontologies After usingdifferent SMs to determine the corresponding similaritymatrices GA is used to optimize their aggregating weights toobtain the final alignment

A security ontology alignmentrsquos quality can be measuredwith metrics in the information retrieval domain [10]

recall |RcapRA|

|R|

precision |RcapRA|

|A|

f minus measure 2 precision middot recallrecall + precision

(1)

where A and RA are respectively an alignment and ref-erence alignment and denotes a setrsquos cardinalities Here f-measure is the harmony mean of recall and precision Onthis basis the security ontology alignment problem has theobjective to maximize the f-measure and the decisionvariable is X (x1 x2 )T where xi isin [0 1] i 1 2 is the ith SMrsquos aggregating weight and 1113936 xi 1 In thiswork we choose the weighted average strategy to aggregatethe SMs which is the most popular and flexible method inthe domain of information fusion of combining SMs -eother aggregating mechanisms such as those in the field ofevidential reasoning and fuzzy reasoning could be alsoapplied which is one of our future works

22 Similarity Measure SM can generally be categorized aseither syntactic linguistic or taxonomy SM [11 12] whichwe describe as follows

Syntactic SM calculates the similarity of two stringsthrough their edit distance We use the Levenshtein distance[13]

Levenshtein s1 s2( 1113857 max0 min s1

11138681113868111386811138681113868111386811138681113868 s2

111386811138681113868111386811138681113868111386811138681113872 1113873 minus d s1 s2( 1113857

min s11113868111386811138681113868

1113868111386811138681113868 s21113868111386811138681113868

11138681113868111386811138681113872 1113873

(2)

where |s1| and |s2| are the respective character numbers ofstrings s1 and s2 and d(s1 s2) is their edit distance

Linguistic SM utilizes an electronic dictionary to mea-sure the similarity of two words We useWordNet [14 15] asthe electronic knowledge base Linguistic similarity is de-fined as

Linguistic w1 w2( 1113857 maxc1isinsen w1( )c2isinsen w2( )

sim c1 c2( 11138571113858 1113859

(3)

where w1 and w2 are words derived from two entities andsen(wi) denotes the number of meanings of wi

SecurityAssertion

Security Protocol SecurityEncryption SecuritySignature Security Algorithm

ProtocolSignature

OpenPGP-SignatureXML-Signature

ProtocolAuthentication

KerberosUsername

ProtocolCanonization

Canonization

ProtocolEncryption

OpenPGP-EncryptionXML-Signature

ProtocolAccessControl

XACMLACL

Component

Elements

Data

AlgAsymmetric

RSADSA

AlgSymmetric

DESAES

AlgHash

SHA1MD5

AlgCanonization

XMLC14N

SecurityMichanism

DigitalDigestDigitalSignature

SecurityToken

UsernameTokenX509CertificateKerberosTickethellip

hellip

hellip

hellip hellip hellip hellip hellip

hellip

hellip

hellip

Figure 1 An example of security ontology




2 (4)




Ontology Alignment

OOprime

e1

e1

e2

e2

ei

ej en

em

OOprime

e1

e1

e2

e2

ei

ej en ej en

em

OOprime

e1

e1

e2

e2

ei

em

02

01

01

02

01 01

03

03

01

02 01

03

02

03

06



hellip




hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

helliphellip

hellip

hellip





wk

c1 k 1


1 minus cn k n + 1

⎧⎪⎪⎨

⎪⎪⎩(5)








end for





end while



1Chromosome

051 019

0 0 0 0 0 0 01 1

Cprime

019 051C

019 051


c1

c2prime

c2

w1 w2 w3

c1prime

c2c1








00102030405060708091

1XX2XX

GA

CroM

atch

LogM

ap

XMap

CroM

atcher

LogM

ap-C

AOT_

2014

MaasM

tch

ORe

ason

er

HerTU

DA

IAMA

MapSSS

RIMIO

M2013

Strin

gsAu

to

WeSeE

XMapGen

YAM++ ASE

GOMMA

Optim

a

ServOMaplt

Wmath






4 Conclusions



Data Availability




Acknowledgments


References






































2 (4)




Ontology Alignment

OOprime

e1

e1

e2

e2

ei

ej en

em

OOprime

e1

e1

e2

e2

ei

ej en ej en

em

OOprime

e1

e1

e2

e2

ei

em

02

01

01

02

01 01

03

03

01

02 01

03

02

03

06



hellip




hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

hellip

helliphellip

hellip

hellip





wk

c1 k 1


1 minus cn k n + 1

⎧⎪⎪⎨

⎪⎪⎩(5)








end for





end while



1Chromosome

051 019

0 0 0 0 0 0 01 1

Cprime

019 051C

019 051


c1

c2prime

c2

w1 w2 w3

c1prime

c2c1








00102030405060708091

1XX2XX

GA

CroM

atch

LogM

ap

XMap

CroM

atcher

LogM

ap-C

AOT_

2014

MaasM

tch

ORe

ason

er

HerTU

DA

IAMA

MapSSS

RIMIO

M2013

Strin

gsAu

to

WeSeE

XMapGen

YAM++ ASE

GOMMA

Optim

a

ServOMaplt

Wmath






4 Conclusions



Data Availability




Acknowledgments


References






































wk

c1 k 1


1 minus cn k n + 1

⎧⎪⎪⎨

⎪⎪⎩(5)








end for





end while



1Chromosome

051 019

0 0 0 0 0 0 01 1

Cprime

019 051C

019 051


c1

c2prime

c2

w1 w2 w3

c1prime

c2c1








00102030405060708091

1XX2XX

GA

CroM

atch

LogM

ap

XMap

CroM

atcher

LogM

ap-C

AOT_

2014

MaasM

tch

ORe

ason

er

HerTU

DA

IAMA

MapSSS

RIMIO

M2013

Strin

gsAu

to

WeSeE

XMapGen

YAM++ ASE

GOMMA

Optim

a

ServOMaplt

Wmath






4 Conclusions



Data Availability




Acknowledgments


References









































00102030405060708091

1XX2XX

GA

CroM

atch

LogM

ap

XMap

CroM

atcher

LogM

ap-C

AOT_

2014

MaasM

tch

ORe

ason

er

HerTU

DA

IAMA

MapSSS

RIMIO

M2013

Strin

gsAu

to

WeSeE

XMapGen

YAM++ ASE

GOMMA

Optim

a

ServOMaplt

Wmath






4 Conclusions



Data Availability




Acknowledgments


References







































4 Conclusions



Data Availability




Acknowledgments


References




































matching cyber security ontologies through genetic

Documents