WIXLIB

Wireless Communications and Mobile Computing3Table ]This study proposes a multistepThe data used in the case study isThough the proposed method seemsapproach to evaluate, categorize, andlimited which explains the flow of thecomplex but a real-timerate cloud-based IoT platforms viamethod but falls back to prove itsimplementing Multicriteria Decision- implementation via case study providescogency. Moreover, inclusion of latestcogent proof of its efficiency. It alsoMaking (MCDM), probabilistichybrid techniques in the domain foroutperforms individual scoring,linguistic term sets (PLTSs), and finally,comparative analysis could further edifyclassification, and evaluating methods.a probabilistic linguistic best-worstthe study’s significance.(PLBW) is used to score all platforms[9]Cloud service provider selectionapproach is proposed via application of The study successfully identifies andprovides solutions the drawbacks ofMulticriteria Decision-MakingThe use case scenario presented usedclassical multicriteria decision-making(MCDM), analytical hierarchicalstimulated scenarios and data that raises(MCDM) approaches in terms ofprocess (AHP), technique for order ofquestion against the cogency of thepreference by similarity to ideal solution accuracy, time required, and complexityproposed study.(TOPSIS), and the best-worst method of computation. AHP is outperformedby proposed approach.(BWM). Case study is presented tosupport the same.[10]Additive manufacturing based cloudThis study understands and providesbased service providing framework isThe study only provides a frameworkproposed to include both hard and soft solution to the real-time consumer oralong with it merit without any detailscustomer problems. Its featureservices for the ease of customer use.of implementing or developing theproviding framework proves to be easy,These include data-based testing,framework for real world application.feasible, and effective.design, 3D printing, remote control ofprinters, and face recognition using AI.[11]The study is aimed at identifyingvarious determinants that causeA comparative analysis with otherdeprecation of various ministry ofThis study evaluates real-time data frompolicy-making insight providermicro, small, and medium enterprises500 MSMEs that proves its cogency.algorithms along with impact ofin India, contributing a huge impact onMoreover, it provides insight that canimplementing the recommendedIndian economy. Data is collated frombe directly deliberated by policy makerschanges would create more clarity and500 Indian MSMEs. Multiple criteriato create maximum impact.value for the research.include social influence, Internet ofThings, perceived ease of use, trust, andperceived IT security risk among others.[12]A comparative analysis is performed toobtain the best cloud-based IoTplatform for any business orThe hierarchal method of requirementorganization by deliberating multipleclassification provides edge to thecriteria, functional and nonfunctionalmethod and various statistical testsrequirements among five giants,implemented on the results obtainednamely, Azure, AWS, SaS, ThingWorx,creates increased sense of cogency orand Kaa IoT by application varioussignificance to the study.techniques like analytical hierarchicalprocess (AHP), K-means clustering,and statistical tests.[13]IoT applications built via cloud-basedplatforms are assayed for any kind ofsecurity challenge or data inconsistencyissues that arise due to third partyauditors, phishing attacks. It alsoprovides strategies to prevent the same.The objective of the study is veryrelevant to the need of the hour,providing valid and much neededinformation. It also providesrecommendations handle the same.The cloud-based IoT platforms arelimited creating false sense ofperformance in terms of evaluatingmore than 5 platforms. Moreover,requirement classification intohierarchy is very time and effortintensive.The scope of study is limited totheoretical analysis without any realdata implementation or case study toprove the cogency of the pointsmentioned in the paper.

4Wireless Communications and Mobile ComputingTable 1: antages[14]The aim of the study very cogent andThis study poses to create a need forcurrent, deliberating recentThe framework proposed for controlauthorization in cloud-enabled IoTdevelopments of cloud-based IoTbased authorization lacks any sense ofsystems by assaying various securityapplications. The case studies presentedimplementation or efforts towardsthreats that such a set up encounters viaaid to the cause of study whileprototype development.two case studies in order. Proposingcontributing to the significance of thecontrol-based authorization systemproposed framework.[15]An attack distribution detector isproposed to prevent malfunctioning ofThe continuous updating andtrust bounders in IoT-basedverification of the model provides it Inclusion of latest hybrid techniques inapplications, leading to severe datathe domain for comparative analysisoptimal results and performance totheft. A downsampler-encoder-basedcould further edify the study’sdetection probable data thefts. Thecooperative data generator is proposedsignificance.model outperforms primordial machineto discriminate noisy data that may leadlearning and deep learning techniques.of data theft that malfunction trustboundaries.[16]Various cogent issues with IoTmiddleware are brought to attentionwhile proposing a state of art IoTmiddleware that can integrate withMQTT, CoAP, and HTTP asapplication-layer protocols.The problem addressed by In.IoTframework is cogent, and its relevancehas been shown very accurately in thestudy.A comparative analysis with classicalmiddleware and latest hybridtechniques could further edify thesignificance of the study.[17]An intrusion detection technique forcloud-based IoT application isproposed by implementing machinelearning, to obtain state of art accuracyand in-depth analysis of source or typeof intrusion.The survey of 95 developments inintelligence-based intrusion detectiontechniques provides the studysignificant relevance and ground forcomparative analysis with proposedtechnique.Though the study shows optimalaccuracy, false-positive results stillhamper its cogency.distribution when transmitting. Malicious IoT users add orremove the distribution layers to hamper the cycle or amountof reward distributed [25].X1OPOptimal pointAlt3.2. Distance-Based Approach. Distance-based approach(DBA) is an effective and efficient MCDM method. Identifying and defining the optimized state of the multiple attributesthat are part of the process is the initial gradation in the proposed method. The optimal state represented by the vectorOP is the set of best values of criteria over a range of alternatives. The best values can be maximum or minimum, defining the type of criteria.Reference to Figure 1, as indicated, vector “OP” is theoptimal point in a multidimensional space. It acts as a reference point to which the other values of all the alternatives areanalyzed to one another quantitatively. In other words, anarithmetical difference of the current values of alternativesfrom their corresponding optimal values is taken, which represents the ability of the considered alternatives to achievethe optimal state. The decision-making issue which needsto be dealt with is searching for a viable solution on basis ofits proximity to the optimal state.In Figure 1, “H” represents the feasible region and “Alt”as the alternative. The distance-based technique is aimed atdetermining a point in the “H” region and is in closest proximity to the optimal point.HFeasible region0X2Figure 1: Graphical Representation of DBA Methodology.To implement the above approach, let i 1, 2, 3, 4 n alternatives, and j 1, 2, 3, 4 m selection attributes. Amatrix is created to represent the entire set of alternativesalong with their respective criteria, which is shown in (1).2d1166 d216½d 66 4dn1d 12d 22 d n2 d 1m37d 2m 77:7 75 d nm ð1Þ

Wireless Communications and Mobile Computing5This matrix is known as the decision matrix [d]. Now, wetake the priority weights of these attributes according to theopinions of various experts and calculate their averages. Wetake the sum of these averages and divide the sum by eachof these averages. The result is the creation of another matrixwith only one row and columns equal to the number of attributes. This matrix is known as priority weights matrix [PW]as shown in (2).½PW ½PW11 PW12 PW1m :StartDecision maker’s team formationSelection attributes identificationð2ÞAlternative identificationUsing the following Equations (3), (4) and (5), the decision matrix is standardized to minimize the impact of different units of measurement and to simplify the process.1 n d ,n i 1 ijdj Weights/ratings estimation by DMð3Þwhere d j is the average of each attribute for all alternatives." 21 n S j d ij d jn i 1#1/2,Selection of best values andformation of optimal matrixð4Þd ij d j,SjCalculation of distance of eachselection attributes andformation of distance matrixð5ÞFormation of weighted distance matrixwhere dij is the value of each attribute for an alternative andd ′ij is the standardized value of each attribute for analternative.The final matrix is known as the standardized matrix [d ′ ]and is represented as in (6).2d′116h i 6 d′21d′ 666 4d ′n1d′12d′22 d ′n2 ReviewYeswhere S j is the standard deviation of each attribute for allalternatives.d′ij NoApprovedd ′1mStopð6ÞThe best value of each attribute is selected over the set ofalternatives. The best values can be maximum or minimumvalues, depending on the type of attribute specified. Thematrix formed using this set of values is known as the optimal matrix [O] as shown in (7).½O ½O11 O12 O1m :Calculation of rank of the alternatives37d ′2m 77:7 75 d ′nm Formation of composite distance matrixð7ÞThe distance of each of the alternatives from its optimalstate is calculated as the numerical difference between thevalues of each of the attributes and their corresponding optimum counterparts. The resulting values form a matrix calledFigure 2: Model development of DBA methodology.the distance matrix [O ′ ] as represented in (8).2h′O11 d 116i 6 O d′1121O′ 666 4′ O12 d 12′ O12 dn2′O11 dn137′ 7O1m d2m7:77 5′ O1m dnm′ O12 d 22 ′O1m d1mð8ÞEach value of this matrix is then squared and multipliedby their corresponding priority weights, as explained byEquation (9).

6Wireless Communications and Mobile gicalfactorsEconomicfactors(i) Functionality(ii) Reliability(iii) Usability(iv) Efficiency(v) Maintainability(vi) Portability(i) Storage capacity(ii) CPU performance(iii) Memoryutilization(iv) Platform design(v) Network speed(i) Serviceinduction cost(ii) Maintenancecost(iii) PromotioncostFigure 3: Classification of selection attributes.Table 2: TFN scale for priority weights.2W ij O′ij PW j :ð9ÞThe resulting matrix is called the weighted distancematrix [W] as shown in (10).2W 1166 W 216½W 66 4W n1W 12W 22 W n2 W 1m37W 2m 77:7 75 W nm ð10ÞSymbolsExtremely most important (EMI)Very most important (VMI)Most important (MI)Important (I)Less important (LI)Very less important (VLI)Extremely less important (ELI)m:ð11Þj 1The one-column matrix formed as a result of this equation is called the composite distance matrix [CD] as shownin (12).2CD113676 CD21 767:½CD 676 745CDn1Table 3: TFN scale of performance ratings.Symbols#1/2CDi W ij[1, 1, 1][0.8, 0.9, 1][0.6, 0.7, 0.8][0.4, 0.5, 0.6][0.2, 0.3, 0.4][0.0, 0.1, 0.2][0.0, 0.0. 0.0]Table of TFN scale for importance of the 14 attributes during cloud serviceprovider selection.Equation (11) is used to calculate the composite distance,“CD” between each alternative to the optimal state."TFN valuesVery high (VH)High (H)Above average (AA)Average (A)Below average (BA)Low (L)Very low (VL)TFN values[1, 1, 1][0.8, 0.9, 1][0.6, 0.7, 0.8][0.4, 0.5, 0.6][0.2, 0.3, 0.4][0.0, 0.1, 0.2][0.0, 0.0. 0.0]Table of TFN scale for performance ratings of the 9 CSPs over the selectionattributes.ð12ÞThe last step of this method involves calculating the rankof each alternative by using their composite distance values.The smallest value gets the 1st rank, the second smallest valuegets the 2nd rank, and so on. This is how the DBA MCDMapproach is used for cloud service provider selection. Below,Figure 2 represents the model development of themethodology.3.3. Estimation-of-Distribution Algorithms. These algorithmsare general metaheuristics applied in optimization to represent a recent alternative to classical approaches [21]. EDAsbuild probabilistic models of promising solutions by repeatedly sampling and selecting points from the underlyingsearch space. EDAs typically work with a population of candidate solutions to the problem, starting with the populationgenerated according to the uniform distribution over alladmissible solutions [22]. Many distinct approaches havebeen proposed for the estimation of probability distribution,

Wireless Communications and Mobile Computing0.60.520.480.38D 0.120.260.520.180.6(a)PW [0.078 0.078 0.074 0.068 0.075 0.071 0.066 0.071 0.075 0.073 0.074 0.062 0.072 0.064 ](b)Figure 4: (a) Decision matrix is represented by D. (b) Performance rating matrix represented by P.1.920.930.43–0.79D′ �1.54–0.750.71–1.201.17Figure 5: Standardized matrix represented by D ′ .O [1.92 1.15 1.28 1.18 2.06 1.21 1.60 1.26 1.50 1.96 1.71 –1.24 –1.74 –1.54](a)00.981.482.71O′ 2.26–0.34–2.72(b)Figure 6: (a) Optimal matrix represented by O. (b) Distance matrix represented by O ′ .such as Independent Variables, Bivariate Dependencies, andMultiple Dependencies.4. Implementation, Results, and DiscussionsEvaluating various cloud service providers using DBA (distance-based approach) methodology with Fuzzy Set Theoryto calculate ranks based upon selection attributes is describedby the following steps:(1) Identification and selection of cloud service providers(CSP): with the aid of an experienced group of dedicated decision-makers, specialized in the field ofComputer Science and Technology, an analysis ofseveral cloud service providers was conducted, and9 most widely used CSPs, namely, (1) Amazon WebServices(AWS), (2) Digital Ocean, (3) Google CloudPlatform, (4) Microsoft Azure, (5) RackSpace, (6)SoftLayer, (7) IBM Smart Cloud Enterprise, (8)GoGrid, and (9) Google Compute Engine, were filtered out by evaluating them on various conceptualand practical criteria based on both quality andusability, thereby selecting them after immensebrainstorming sessions and collective use of elimination principle(2) Identification of selection attributes: three major factors were identified which were, namely, quality factors, technical factors, and economic factors. Theywere further classified as: quality factors (functionality, reliability, usability, efficiency, maintainability,and portability), technical factors (storage capacity,CPU performance, memory utilization, platformdesign, and network speed), and economic factors(service induction cost, maintenance cost, and promotion cost) after the detailed analysis and intensivestudy of the cloud service providing industry and itsvarious prerequisites along with understanding themarket where this industry thrives (Figure 3).(3) Application of Fuzzy Set Theory: fuzzy logic forms itsbasis upon human acumen of decision-makingregarding vague and nebulous information. It grossly

8Wireless Communications and Mobile Computing00.070.170.57W .852CD 1.8071.8001.8501.6242.085(b)Figure 7: (a) Weighted distance matrix represented by W. (b) Composite distance matrix represented by CD.Table 4: Rank of cloud service providers.Cloud service providersAmazon Web ServicesDigital OceanGoogle Cloud PlatformMicrosoft AzureRackSpaceSoftLayerIBM Smart Cloud EnterpriseGoGridGoogle Compute EngineComposite 6242.085825743619(4) Determination of weights and performance ratings:the expert-assigned linguistic terms were first converted into corresponding TFNs using the fuzzy scaleand then defuzzied to get crisp score values. The datawas extracted from the questionnaires and then evaluated using a combination of the mathematical formulas and concepts of aggregation and average(5) Creating performance rating matrices: a decisionmatrix of the performance ratings (Figure 4(a)) and asingle-row matrix of the priority weights (Figure 4(b))were created under the supervision of expert guidanceusing the fuzzy scale and MCDMTable for ranking of various cloud service providers.distinguishes real-world problems based uponhuman comprehensive skills rather than absoluteBoolean logic. In other words, the fuzzy systemimplements scales rather than 0/1 for coherenthuman understanding where 0 represents absolutefallacy, 1 represents absolute truth, and the middlevalues represent the fuzziness or the fuzzy values. Inthis study, we have implemented a triple fuzzy number scale which uses a triplet set of the form [a, b, c]with a sensory scale (Tables 2 and 3) [28]. A surveywas conducted among a group of 40 selected expertsassociated with the technical field. The (Table 2)questionnaire consisted of 14 pristine questions,based upon which a priority weights matrix was created consisting of the weights or values of theassorted attributes. While in the second questionnaire (Table 3), the nine already selected CSPs wereappraised on the grounds of the 14 categorized selection attributes by an adept team of 5 experts. Theextracted data from the questionnaires mentionedabove were converted from a literal scale to TFN(Triple Fuzzy Number) scale, thereby averaged to afuzzy number(6) Calculating standardized matrix: root mean square ofeach selection attributes is carefully evaluated; furthermore, the mean previously determined is subtractedfrom each value and simultaneously divided by the corresponding root mean square of that particular selection attribute to get the standardized matrix (Figure 5).(7) Creating optimal and distance matrix: the optimalmatrix is estimated by targeting the best values ofeach selected attribute of the standardized matrix(Figure 6(a)), i.e., the maximum values for qualityand technical factors and minimum values for economic factors. Additionally, the distance matrix iscalculated by finding the distance between each valueof a particular

To search the relevant data, the keywords like Cloud Computing for IoT, Cloud Platforms for IoT Services, IoT based Cloud Service Selection, IoT Service Selection Attri-butes, and Cloud Service Selection for IoT were used. . Azure, AWS, SaS, ThingWorx, and Kaa IoT by application various techniques like analytical hierarchical process (AHP), K .