Inter national J our nal of Electrical and Computer Engineering (IJECE) V ol. 8, No. 1, February 2018, pp. 304 325 ISSN: 2088-8708 304       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     Pr efer ences Based Customized T rust Model f or Assessment of Cloud Ser vices Shilpa Deshpande and Rajesh Ingle Department of Computer Engineering, Colle ge of Engineering Pune, Sa vitribai Phule Pune Uni v ersity , India Article Inf o Article history: Recei v ed May 13, 2017 Re vised No v 23, 2017 Accepted Dec 7, 2017 K eyw ord: Cloud computing Customized trust assessment Dynamic trust Elastic trust computation Quality of Service (QoS) ABSTRA CT In cloud en vironment, man y functionally similar cloud services are a v ailable. But, the ser - vices dif fer in Quality of Service (QoS) le v els, of f ered by them. There is a di v ersity in user requirements about the e xpected qualities of cloud services. T rust is a measure to under - stand whether a cloud service can adequately meet the user requirements. Consequently , trust assessment plays a significant role in selecting the suitable cloud service. This pa- per proposes preferences based customized trust model (PBCTM) for trust assessment of cloud services. PBCTM tak es into account user requirements about the e xpected quality of services in the form of preferences. Accordingly , it performs customized trust assessment based on the e vidences of v arious attrib utes of cloud service. PBCTM enables elastic trust computation, which is responsi v e to dynamically changing us er preferences with time. The model f acilitates dynamic trust based periodic selection of cloud services according to v ary- ing user preferences. Experimental results demonstrate tha t the proposed preferences based customized trust model outperforms the other model in respect of accurac y and de gree of satisf action. Copyright c 2018 Institute of Advanced Engineering and Science . All rights r eserved. Corresponding A uthor: Shilpa Deshpande, Department of Computer Engineering, Colle ge of Engineering Pune, Sa vitribai Phule Pune Uni v ersity , Pune, Maharashtra, India Email: shilpshree@yahoo.com 1. INTR ODUCTION Cloud computing has entered mainstream and recei v ed wider acceptance. It is increasingly adopted by indi- viduals, small and medium scale enterprises (SMEs) and go v ernment or g anizations to run their critical applications. The reason for this acceptance is the characteristics of cloud lik e scalability , on demand service, an ytime-an ywhere access, economic benefits of pay-per -use, dele g ation of maintenance and administration, performance and disaster re- co v ery . Cloud services ha v e proliferated to include softw are as a service, database as a service, platform as a service, infrastructure as a service, security as a service and storage as a service [1]. Cloud en vironment still remains chal- lenging to rely on because of f actors lik e loss of control o v er applications and data, increased threats of security [2], performance is sues related to virtualization [3], enterprise grade a v ailability requirements [4, 5, 6] and adequately meeting Quality of Service (QoS) e xpectations of users [7]. Cloud computing has compelling adv antages yet challenges too. F or an enterprise to adopt cloud, it is impor - tant that enterprise has a certain belief that adv antages of cloud can be realized. T rust is a measure of this belief [8]. Con v entionally , people rely on reputation [4, 8], service le v el agreement (SLA) [6, 9], self-assessment [8, 9] and cloud auditing [8, 9] for trust assessment in cloud en vironment. Ho we v er , trust assessment in cloud en vironment poses further important issues, which are re v ealed as part of the follo wing discussion. Reputation based traditional trust assessment technique relies on the opinions of cloud users. The opinions tak en in the form of ratings or feedbacks may be subjecti v e in nature [8]. Therefore, reputation cannot be an e xact reflection of realistic capabilities of the cloud service. A service le v el agreement (SLA) established between a cloud service consumer and a pro vider consists of functional and QoS f acets of the of fered service [10]. Le v els of SLA are not consistent among the cloud service pro viders of fering analogous services. Moreo v er , for a service pro vider , J ournal Homepage: http://iaescor e .com/journals/inde x.php/IJECE       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     DOI:  10.11591/ijece.v8i1.pp304-325 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 305 promises made in SLA and actual QoS deli v ered are not consistent. Consequently , it mak es hard for consumers , to e v aluate the trust of a cloud service, solely based on the SLA [6, 9]. Cloud service pro vider may announce the self- assessment of the of fered cloud services, based on cloud t ransparenc y mechanisms [8]. Ho we v er , such e v aluat ion of cloud services reflects merely a generalized trust assessment of cloud services from the vie wpoint of pro vider . The mechanism of self-assessment does not tak e into consideration cloud user’ s perspecti v e. A formal cloud audit based trust e v aluation is an anot h e r method which pro viders may use to ensure the quality of of fered services. Ho we v er , audit report typically represents only a static trust assessment of the service at the time when auditing is done [9]. Cloud QoS attrib utes such as performance, a v ailability , reliability and security are significant for user and hence for trust assessment of a cloud service [9]. P ast recorded e vidences of QoS attrib utes signify the actual v alues and the y represent the realistic capabilities of a cloud service [8]. Therefore, the e vidences of cloud QoS attrib utes are needed to be tak en into account by t he trust e v aluation mechanism. There is a di v ersity in requirements of cloud users about the e xpected qualities of cloud services. Hence, customized trust assessment [9] of a cloud service which tak es into consideration the user preferences for the cloud QoS attrib utes, is needed to enable the personalized selection of suitable cloud service [11]. Cloud service pro vider’ s capacity to pro vide services v aries with time. As a result, the QoS le v els of of fered services also change with time. Therefore, trust e v aluation based on one-time e vidences of QoS attrib utes is not enough and it has to be a cons tant dynamic process [8]. Requirements of the user about the e xpected QoS may change dynamically with time [12]. Cloud service pro vider’ s ability to meet user requirements is not al w ays constant. Therefore, trust assessment which includes one-time checking of requirements of the user is not adequate. Hence, trust assessment needs to be respons i v e to the changes in requirements of the user . This implies the need for elastic trust assessment as per the changes in requirements of the user . In this paper , we present preferences based customized trust model (PBCTM), addressing the abo v e men- tioned issues. More specifically , the contrib utions are: 1. A no v el method for trust comput ation of a cloud service based on the distances of v arious service e vidences from the user preferences. 2. Customized trust assessment mechanism containing mathematical formulation of weights which are computed based on the relati v e importance of cloud service attrib utes with respect to QoS e xpectations of user . 3. Introduction of the concept of elastic trust computation of a cloud service and an algorithm for it. 4. Mechanism for ranking of cloud services based on trust computation which is dynamic, elas tic and considers preferences of users. 5. Comparison of the proposed trust model with other model with re g ard to accurac y and a ne w measure of de gree of satisf action of trust assessment. The paper is or g anized as follo ws. Section 2 presents a re vie w of related w ork. In Section 3, the architecture of the system meant for the proposed trust model and the functional o v ervie w of trust assessment are described. Section 4 defines the preferences based customized trust model (PBCTM) and presents the details of customized and dynamic trust assessment. Section 5 presents the algorithm for elastic trust computation of a cloud service. Section 6 depicts the method for ranking of cloud services based on the proposed trust model. Section 7 presents the qualitati v e comparison of PBCTM with other models. Section 8 co v ers the performance e v aluation of the proposed trust model including the results and analysis. Section 9 concludes the paper . 2. RELA TED W ORK Reputation based approaches mak e use of feedbacks from man y cloud users to e v aluate trust of a cloud service. T rust assessment approaches proposed by [13, 14, 15] are based on reputation. These approaches do not tak e into account requirements of user for t rust e v aluation. Moreo v er , these re p ut ation based approaches f all short in performing dynamic assessment of trust [8]. Besides user feedbacks, fe w of the approaches in literature, tak e into consideration additional f actors such as pro vider’ s self-declarations and e xpert’ s ratings, for trust assessment [16]. Ho we v er , credibility [4] of the f actors included in trust e v aluation is a main concern in these approaches. Habib et al. [10] proposed an architecture to enable trust assessment of cloud service pro viders using v arious f actors such as pro vider statements, user feedbacks and certificates. A trust model based on service le v el agreement (SLA) parameters is proposed by P a w ar et al. [17]. Ghosh et al. [18] proposed a frame w ork for assessment of risk of interaction with cloud service pro vider . The approach in turn includes e v aluating t rust of the service pro vider . The trust is estimated on the basis of direct and indirect interactions Pr efer ences Based Customized T rust Model for Assessment ... (Shilpa Deshpande) Evaluation Warning : The document was created with Spire.PDF for Python.
306 ISSN: 2088-8708 between customer and cloud pro vider . The approaches [10, 17, 18] do not of fer dynamic trust update along a period of time. Also, these approaches do not consider QoS requirements of us er for trust assessment. A model is recommended by Mo yano et al. [19] to e v aluate trust of cloud pro viders . Although, the approach is simple, trust assessment mainly depends on the accessibility to the information released by the cloud pro viders. Fe w of the approaches do tak e into account QoS attrib utes for trust e v aluation. The approach proposed by Manuel et al. [20] e v aluates the trust of a cloud resource in terms of summation of v alues ass igned to user feedbacks, security le v el and reputation. A model is suggested by Manuel et al. [21] to compute the reputation based trust of a resource. The model mak es use of identity , capability and beha vior v alues of a resource to obtai n its trust v alue. The approaches [20, 21] do not consider requirements of users in trust estimation of resources. Also, these approaches do not reflect dynamic trust assessment of resources along a period of time. A fuzzy trust e v aluation approach for cloud services is suggested by Huo et al. [22]. The approach tak es into consideration a set of cloud s ervice attrib utes to assess the reputation based trust v alue. F an et al. [ 23 ] suggested a mechanism for e v aluating dynamic trust of a cloud service using multiple attrib utes. The mechanism of trust computation relies on the feedbacks gi v en by the users. Ho we v er , authenticity of feedbacks is not addressed by the authors. The approach f acilitates selection of a service according to the user requirements for v arious attrib utes. Ho we v er , the approaches [21, 22, 23] are dependent on subjecti v ely allocated weights to the v arious f actors. The QoS based mechanisms in literature, mak e us e of a v ailability , performance, security and reliability as the general attrib utes of cloud service for trust assessment. Throughput, response time, netw ork bandwidth and ca- pability are the usually considered performa nce related f actors in trust estimation. Li et al. [24] proposed a method for dynamic trust e v aluation of cloud resources. It mak es use of recorded v alues of v arious attrib utes for computation of trust. The authors do not focus on consideration of user requirements for attrib utes, in e v aluating trust v alue of a resource. Frame w orks are proposed by [25, 26] for trust e v aluation of cloud service pro viders based on QoS attrib utes. The approaches are based on monitoring QoS attrib utes and e v aluating the compliance with re g ard to the SLA. System suggested by [26] i n c orporates perspecti v es of dif ferent entities such as cloud users, auditor and peers in the process of trust e v aluation. Supriya et al. [27] proposed to emplo y multi-criteria based d e cision making methods for e v aluating trust of cloud service pro viders. The w ork f acilitates ranking of pro viders based on their trust v alues. It of fers per - sonalized computation of trust by considering priorities for the v arious attrib utes of cloud pro vider . Ho we v er , priority based weights assigned to the dif ferent attrib utes are static and subjecti v e. System is proposed by Qu and Buyya [11] for trust estimation of a cloud service based on its performance in terms of v arious QoS attrib utes. The approach tak es into account QoS requirements of user and computes the trust of a service based on fulfillment of the requirements. Ho we v er , the approaches [11, 25, 26, 27] do not of fer dynamic trust update in cloud en vironment. A model is proposed by Manuel [28] to e v aluate trust of a resource based on its capabilities and measured QoS attrib utes. T rust update is indicated only by algorithmic steps. The model enables matching the QoS requirements of users to the resources according to their com p ut ed trust v alues. Ho we v er , static weights based on pre-decided priorities are assigned to the v arious attrib utes. In summary , consideration of user requirements in trust assessment is essential to enable the personalized selection of appropriate cloud services. Ho we v er , the abo v e re vie w of the related w ork signifies that only fe w of the approaches [11, 23, 27, 28] consider requi rements of cloud users for trust assessment. Cloud QoS attrib utes are sig- nificant for trust e v aluation of a cloud service. Evidences of QoS attrib utes obtained through monitoring are unbiased in nature and are more dependable f actors for trust estimation. Ho we v er , the approach [23] does not tak e into account e vidences of QoS attrib utes and trust assessment solely relies on the feedbacks of users. Dynamic cloud en vironment implies the need for trust to be assessed continuously with ti me. Ho we v er , the approaches [11, 27] do not of fer dynamic trust e v aluation of cl oud services. Although, the approach [28] tak es into consideration requirements of users, weights calculation for v arious attrib utes in trust assess ment does not reflect preferences of users. Moreo v er , assessment of trust according to the dynamical ly changing requirements of the user , is not addressed by an y of the abo v e approaches. Our trust model PBCTM, aims to address these limitations in the earlier w ork. PBCTM performs customized trust assessment of a cloud service by taking into account e vidences of service attrib utes and preferences of user for at- trib utes. Our model f acilitates elastic trust computation of a cloud service according to the dynamically changing user preferences of attrib utes with time. PBCTM enables computation of weights for the multiple attrib utes of a service by considering the relati v e utility of attrib utes with respect to the user preferences. Dynamic trust prediction used in our model, allo ws ranking of cloud services to assist the user in periodic selection of suitable service. 3. ARCHITECTURE OF TR UST ASSESSMENT SYSTEM Figure 1a sho ws the o v erall layout of the system meant for the proposed trust model. It depicts the main trust assessment and ranking module which is connected with the other supplementary modules. The functional specification collector compiles the functional requirements of the cloud service, submitted by cloud us er . Multiple IJECE V ol. 8, No. 1, February 2018: 304 325 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 307 service pro viders re gister their services i n t o the service repository . Services Extraction module finds the services from service repository whose functional specifications match with the required one. The user preferences collector compiles the preference v alues for cloud service attrib utes such as a v ailability , throughput and response time, which are submitted by the cloud user . (a) Architecture (b) Functional o v ervie w Figure 1. T rust assessment system for preferences based customized trust model (PBCTM) T rust assessment and ranking module is the core component performing dynamic and elastic trust computa- tion of the cloud services. F or each of the matching cloud services, the trust assessment is carried out by taking into account the user preferences and the e vidences of service attrib utes. The results of trust assessment and ranking are recorded in the customized trust archi v es. The cloud user can select the appropriate cloud s ervice based on the ranking of cloud services. The process of monitoring, continuously observ es and records the v alues of attrib utes such as response time, throughput, a v ailability and security , for each of the cloud services. The e vidence collector collects the e vidence f actors, recorded as part of continuous monitoring process. These e vidence f actors are then used for trust assessment of cloud services. T rust assessment and ranking is the main focus of this paper . Hence, the details of other modules which include services e xtraction, monitoring and related functionalities, are not discussed further , in this paper . W e assume these as the already e xisting v alid services and are a v ailable in the form of e xternal interf aces to the trust assessment and ranking module. Figure 1b sho ws the high-le v el functional o v ervie w for trust assessment and ranking of cloud services. The user preferences for v arious service attrib utes, are tak en as input for the trust assessment. Evidence f actors for each of the matching cloud services, o v er the period of time, are tak en as another input by the trust assessment module. The module calculates customized present trust of services at an instant of time, by considering the user preferences and the corresponding service e vidence f actors. Subsequently , the module performs dynamic prediction of trust v alues of cloud services o v er a period of time. The customized present trust and predicted trust v alues are returned to the cloud user . The ranking of cloud services is performed by the module, based on the predicted trust v alues of the services. The resultant ranking sequence of cloud services, which is then returned to the cloud user , f acilitates the customized selection of suitable cloud service for the user . The preferences of a particular cloud user may change dynamically with time. Accordingly , the operations of trust assessment and ranking of cloud services are performed repetiti v ely with changing preferences of the user and the continuous e vidences of each cloud service. This reflects the dynamic and elastic trust computation of cloud services. The cloud user can re vise the selection of a suitable service based on the updated ranking of cloud services. The details of customized and dynamic trust asses sments of a cloud service are described in Section 4. The steps depicting the control flo w for elastic trust computation are presented i n Section 5 in the form of algorithm. The trust based ranking of cloud services is elaborated in Section 6. 4. PREFERENCES B ASED CUST OMIZED TR UST MODEL T rust assessment of a cloud service is performed based on the preferences of a cloud user for service attrib utes and the e vidence f actors of the service. Evidence f actors of a cloud service signify the recorded v alues of service attrib utes. Pr efer ences Based Customized T rust Model for Assessment ... (Shilpa Deshpande) Evaluation Warning : The document was created with Spire.PDF for Python.
308 ISSN: 2088-8708 Definition 1 Pr efer ences Based Customized T rust Model (PBCTM) is defined by a 12-tuple: ( L; AC ; T I ; P R ; C ; N C ; P D P ; N D P ; C P T ; C T ; E ; D ) wher e L: Set of v cloud services: f s 1 ; s 2 ; :::; s v g A C: Set of m cloud service attrib utes: f R 1 ; R 2 ; :::; R m g TI: Or der ed discr ete set of n time instances, in a time window: f 1 ; 2 ; :::; n g PR: Set of pr efer ences of a user for the values of cloud service attrib utes: f pr 1 ; pr 2 ; :::; pr m g C: An e vidence matrix whic h depicts m e vidence factor s at eac h of the n time instances. NC: Normalized augmented e vidence matrix with pr efer ences. PDP: Normalized matrix for positive distances fr om pr efer ences. NDP: Normalized matrix for ne gative distances fr om pr efer ences. CPT : Customized Pr esent T rust of a cloud service at a particular time instant. CT : Cumulative T rust of a cloud service o ver a period of time . E: A set of cor e trust assessment functions: f f C P T ; f C T g ; wher e f C P T indicates a function to compute Customized Pr esent T rust (CPT) and f C T is a function to assess Cumulative T rust (CT). D: A set of allied functions: f f N E ; f P D ; f N D ; f C W g ; wher e f N E is a function to normalize e vidence factor s and pr efer ences; f P D and f N D ar e the functions to compute summative positive and ne gative distances fr om pr efer ences; f C W indicates a function to compute weights of cloud service attrib utes. Evidence f actors of a cloud ser vice are retrie v ed after e v ery fix ed time interv al. Representation of the e vidence f actors is sho wn by an e vidence matrix as: C = 2 6 6 6 4 c 11 c 12 : : : c 1 m c 21 c 22 : : : c 2 m . . . . . . . . . . . . c n 1 c n 2 : : : c nm 3 7 7 7 5 (1) In Equation (1), at a particular time instant i in a time windo w , such that 1 i n , a ro w in the matrix indicates a sample of e vidence f act ors as f c i 1 ; c i 2 ; :::; c im g and each v alue c ij in the sample, de n ot es a v alue of an attrib ute R j . Thus there are n samples of e vidence f actors. Column position in the matrix indicates a specific attrib ute within the sample. Preferences for the v alues of cloud service attrib utes, as specified by the user , are combined with the original e vidence matrix, to obtain the augmented matrix, as sho wn belo w . C P = 2 6 6 6 6 6 4 c 11 c 12 : : : c 1 m c 21 c 22 : : : c 2 m . . . . . . . . . . . . c n 1 c n 2 : : : c nm pr 1 pr 2 : : : pr m 3 7 7 7 7 7 5 (2) In Equation (2), the last ro w in the matrix indicates a sample of preferences as f pr 1 ; pr 2 ; :::; pr m g and each v alue pr j in the sample denotes a preference v alue of an attrib ute R j . F or a cloud service, higher v alues for attrib utes such as a v ailability and throughput are desired. Whereas, lo wer v alues for attri b utes such as response time and security violation incidents are e xpected. If the preference v alue for an y of the attrib utes is not specified by the user , then it reflects that, a minimum quality le v el for that service attrib ute is acceptable to the user . Hence, in such case the preference v alue for the attrib ute in matrix C P is set to a minimum or maximum v alue of the service attrib ute in the time windo w , based on the higher -v alue type of attrib ute (e. g. a v ailability) or the lo wer -v alue type of the attrib ute (e. g. response time), respecti v ely . In order to transform all the v alues in matrix C P to uniform range and to mak e them independent of units, v alues of the matrix C P need to be normali zed. Normalization includes scaling of the v alues. Thus, for further processing of distance computation, each v alue in the matrix C P is normalized in the range denoted by [ R new min ; R new max ] . From the perspecti v e of desired performance of a cloud service, attrib utes can be cate gorized in tw o types: one where higher v alues of an attrib ute R j are desired and the other where lo wer v alues of R j are desired. The cate gory where higher v alues of R j are desired, the corresponding normalized v alues x ij and y j for c ij and pr j respecti v ely , are formulated as: x ij = ( c ij R min j )( R new max R new min ) ( R max j R min j ) + R new min (3) y j = ( pr j R min j )( R new max R new min ) ( R max j R min j ) + R new min (4) IJECE V ol. 8, No. 1, February 2018: 304 325 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 309 The other cate gory where lo wer v alues of R j are desired, the corresponding normalized v alues x ij and y j for c ij and pr j respecti v ely , are de vised as: x ij = ( R max j c ij )( R new max R new min ) ( R max j R min j ) + R new min (5) y j = ( R max j pr j )( R new max R new min ) ( R max j R min j ) + R new min (6) In Equations (3) to (6), R min j is the minimum v alue of the attrib ute R j and R max j is the maximum v alue of R j in matrix C P . The normalized augmented matrix is: N C = 2 6 6 6 6 6 4 x 11 x 12 : : : x 1 m x 21 x 22 : : : x 2 m . . . . . . . . . . . . x n 1 x n 2 : : : x nm y 1 y 2 : : : y m 3 7 7 7 7 7 5 (7) In normalized matrix N C , greate r v alue for an y service attrib ute R j where 1 j m , indicates a higher quality of a cloud service than the quality of a cloud service corresponding to the lo wer v alue of the attrib ute. 4.1. Computation of Distances fr om Pr efer ences If the v alues of service attrib utes are higher as compared to the corresponding preference v alues, it reflects a more trustw orthiness of a cloud service. Here we introduce, the ne w terms, Positi v e Distance ( P D ) and Ne g ati v e Distance ( N D ) to define the comparison of the service attrib ute v alues and the associated preference v alues. P D and N D are the measures for assessment of ho w closely a cloud service meets or f ails to meet the user e xpectations. Definition 2 P ositive Distance ( P D ) and Ne gative Distance ( N D ) for any value a ij in matrix N C , wher e 1 i ( n + 1) , of attrib ute R j fr om its corr esponding pr efer ence value y j , ar e formulated as shown in T able 1. T able 1. Distances from Preferences Scenarios for a ij P D N D Attrib ute v alue ( a ij ) = Preference v alue ( y j ) y j 0 Attrib ute v alue ( a ij ) > Preference v alue ( y j ) y j + a ij y j a ij Attrib ute v alue ( a ij ) < Preference v alue ( y j ) a ij y j y j a ij As defined in T able 1, if the attrib ute v alue is greater than or equal to the preference v alue, then its ( P D ) is higher than its ( N D ) . If the attrib ute v alue is lesser than the preference v alue, then its ( P D ) is lesser than its ( N D ) . Also, when the attrib ute v alue is greater than the preference v alue, then: i) its ( P D ) is higher than the ( P D ) for an attrib ute whose v alue equals the preference v alue. ii) its ( N D ) is lesser than the ( N D ) for an attrib ute whose v alue equals the preference v alue. Whereas, when the attrib ute v alue is lesser than the preference v alue, then: i) its ( P D ) is lesser than the ( P D ) for an attrib ute whose v alue equals the preference v alue. ii) its ( N D ) is higher than the ( N D ) for an attrib ute whose v alue equals the preference v alue. Thus, for the v alues of all the attrib utes in matrix N C , which include both, normalized service e vidence f actors and preference v alues, computations of P D and N D v alues are performed. The computed P D and N D v alues are represented in the form of Positi v e Distance and Ne g ati v e Distance matrices respecti v ely , as: P S = 2 6 6 6 6 6 4 ps 11 ps 12 : : : ps 1 m ps 21 ps 22 : : : ps 2 m . . . . . . . . . . . . ps n 1 ps n 2 : : : ps nm ps ( n +1)1 ps ( n +1)2 : : : ps ( n +1) m 3 7 7 7 7 7 5 (8) Pr efer ences Based Customized T rust Model for Assessment ... (Shilpa Deshpande) Evaluation Warning : The document was created with Spire.PDF for Python.
310 ISSN: 2088-8708 N S = 2 6 6 6 6 6 4 ns 11 ns 12 : : : ns 1 m ns 21 ns 22 : : : ns 2 m . . . . . . . . . . . . ns n 1 ns n 2 : : : ns nm ns ( n +1)1 ns ( n +1)2 : : : ns ( n +1) m 3 7 7 7 7 7 5 (9) In Equation (8), at position i such that 1 i n , a ro w in the ma trix P S indicates a sample of positi v e distances as f ps i 1 ; ps i 2 ; :::; ps im g corresponding to e vidence sample f x i 1 ; x i 2 ; :::; x im g of matrix N C . Here, ps ij denotes a P D v alue for an e vidence f actor x ij of an attrib ute R j from its preference v alue y j . Similarly , in Equation (9), at position i such that 1 i n , a ro w in the matrix N S indicates a sample of ne g ati v e distances as f ns i 1 ; ns i 2 ; :::; ns im g corresponding to e vidence sample f x i 1 ; x i 2 ; :::; x im g of matrix N C . Here, ns ij denotes a N D v alue for an e vidence f actor x ij of an attrib ute R j from its preference v alue y j . The ( n + 1) th ro ws in matrices P S and N S represent the samples of positi v e and ne g ati v e distances respecti v ely , for the sample f y 1 ; y 2 ; :::; y m g of preferences in matrix N C . F or ne xt processing of customized trust computation, all the distance v alues in the matrices P S and N S are normalized in the range denoted by [ D new min ; D new max ] . This con v ersion of all the distance v alues to uniform range is made by preserving the original relati v e ordering among the distance v alues for each of the attrib utes. F or each v alue ps ij in matrix P S , where 1 i ( n + 1) , the normalized v alue pd ij is formulated as sho wn belo w . pd ij = ( ps ij P min j )( D new max D new min ) ( P max j P min j ) + D new min (10) where P min j is the minimum v alue of positi v e distance and P max j is the maximum v alue of positi v e distance for attrib ute R j in matrix P S . The normalized positi v e distance matrix is: P D P = 2 6 6 6 6 6 4 pd 11 pd 12 : : : pd 1 m pd 21 pd 22 : : : pd 2 m . . . . . . . . . . . . pd n 1 pd n 2 : : : pd nm pd ( n +1)1 pd ( n +1)2 : : : pd ( n +1) m 3 7 7 7 7 7 5 (11) F or each v alue ns ij in matrix N S , where 1 i ( n + 1) , the normalized v alue nd ij is formulated as sho wn belo w . nd ij = ( ns ij G min j )( D new max D new min ) ( G max j G min j ) + D new min (12) where G min j is the minimum v alue of ne g ati v e distance and G max j is the maximum v alue of ne g ati v e distance for attrib ute R j in matrix N S . The normalized ne g ati v e distance matrix is: N D P = 2 6 6 6 6 6 4 nd 11 nd 12 : : : nd 1 m nd 21 nd 22 : : : nd 2 m . . . . . . . . . . . . nd n 1 nd n 2 : : : nd nm nd ( n +1)1 nd ( n +1)2 : : : nd ( n +1) m 3 7 7 7 7 7 5 (13) 4.2. Distance based Calculation of Customized Pr esent T rust Customized present trust of a cloud service is an indication of relati v e quality of the service at an instant of time, with re g ard to the e xpectations of the user . Hence, for ef fecti v e customized trust assessment of a cloud service, e vidence f actors need to be e v aluated on the basis of their positi v e and ne g ati v e distances from the preference v alues. Consequently , all the m positi v e distance v alues in sample i such that 1 i ( n + 1) , of matrix P D P , are aggre g ated based on weights of attrib utes, to form a summati v e measure of positi v e distances, as sho wn belo w . S P i = X m j =1 w j pd ij (14) where pd ij is a normalized positi v e distance for attrib ute R j in sample i . Similarly , all the m ne g ati v e distance v alues in sample i such that 1 i ( n + 1) , of matrix N D P , are aggre g ated based on weights of attrib utes, to form a IJECE V ol. 8, No. 1, February 2018: 304 325 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 311 summati v e measure of ne g ati v e distances, as sho wn belo w . S N i = X m j =1 w j nd ij (15) where nd ij is a normalized ne g ati v e distance for attrib ute R j in sample i . In Equations (14) and (15), w j is a weight assigned to cloud service attrib ute R j such that 0 < w j < 1 and P m j =1 w j = 1 . Static weights are not suitable for ef fecti v e customized trust assessment of a cloud service. Hence, weights are needed to be computed by taking into consideration the user preferences for v arious attrib utes of a cloud service. The details of computation of weights for v arious cloud service attrib utes, are described in Section 4.3. F or an e vidence sample at time instant i such that 1 i n , corresponding S P i from Equation (14) indicates a weighte d sum of positi v e dis tances of a ll m e vidence f actors and correspondi n g S N i from Equat ion (15) indi cates a weighted sum of ne g ati v e distances of all m e vidence f actors in the sample. When v alues of e vidence f actors of a cloud service match the user preferences, then it indicates a good cloud service in terms of mee ting the user e xpectations. If positi v e distances of e vidence f actors are higher than the corresponding ne g ati v e distances, then the cloud service meets the requirements of the user . Consequently , for an e vidence sample at time instant i such that 1 i n , higher v alue of summati v e positi v e distance ( S P i ) signifies the better trustw orthiness of a cloud service. Therefore, customized present trust of a cloud service at time instant i , is formulated as a relati v e share of summati v e positi v e distance ( S P i ) o v er S P i and S N i . Definition 3 Customized T rust value of a cloud service ( s l ), at a time instant i , termed as Customized Pr esent T rust (CPT) is defined as: C P T i ( s l ) = S P i S P i + S N i (16) wher e S P i is a summative positive distance and S N i is a summative ne gative distance of e vidence factor s for all the m attrib utes of the service , in sample i suc h that 1 i n and 0 < C P T i ( s l ) < 1 . 4.3. Computation of W eights W eight assigned to an attrib ute signifies the import ance of the attrib ute in trust calculation. W eight of an attrib ute is computed based on the relati v e utility of the attrib ute with respect to preference v alue of the attrib ute. Definition 4 Utility de gr ee of an attrib ute R j , in a time window containing n e vidence samples, is formulated as given below . U ( R j ) = ( n X i =1 x ij ) =y j (17) wher e x ij is a normalized e vidence factor of attrib ute R j at time instant i and y j is a corr esponding normalized pr efer ence value for the attrib ute . From Equation (17), when all the e vidence f actors of an attrib ute in a t ime windo w of size n , e xactly match with the preference v alue, utility de gree of the attrib ute becomes equal to n . When one or more e vidence f actors are less than the speci fied preference v alue, utility de gree of the attrib ute reduces to a v alue which is less than n . Whereas, when utility de gree of the attrib ute goes be yond n , it implies that one or more e vidence f actors are greater than the preference v alue. This is the most desirable situation, where the cloud service attrib ute meets the e xpected quality requirements. Thus, the v alues of utility de gree for v arious attrib utes within a sample, signify the proportionate ef fect on the weights of cloud service attrib utes. Accordingly , weight w j of an attrib ute R j is computed as sho wn belo w . w j = U ( R j ) = m X k =1 U ( R k ) (18) where 0 < w j < 1 and P m j =1 w j = 1 . Higher is the utility de gree U ( R j ) of the attrib ute, greater is its resultant weight. The weights computed using Equati on (18), are substituted in Equations (14) and (15), which subsequently results in customized trust estimation of a cloud service, from Equation (16). Pr efer ences Based Customized T rust Model for Assessment ... (Shilpa Deshpande) Evaluation Warning : The document was created with Spire.PDF for Python.
312 ISSN: 2088-8708 4.4. Calculation of Thr eshold T rust The preferences for v arious attrib utes are in turn used to deri v e the minimum e xpected trust v alue for a cloud service. This trust v alue is termed as threshold trust v alue, which serv es as a baseline with which computed trust v alues can be compared. From Equation (14), S P ( n +1) indicates a weighted sum of positi v e distances of preference v alues of all m attrib utes and corresponding S N ( n +1) from Equation (15), represents a weighted sum of ne g ati v e distances of preference v alues of all m attrib utes. On the lines of C P T in Equation (16), a threshold trust v alue of a cloud service at the specified preferences, is formulated as a relati v e share of S P ( n +1) o v er S P ( n +1) and S N ( n +1) , as sho wn belo w . T pr ( s l ) = S P ( n +1) S P ( n +1) + S N ( n +1) (19) 4.5. Pr ediction of Cumulati v e T rust fr om Customized Pr esent T rust A set of customized present trust ( C P T ) v alues computed at dif ferent time instances forms a time series. From Equation (16), at time instant n , time series ( C T S ) is: C T S = f C P T 1 ( s l ) ; C P T 2 ( s l ) ; :::; C P T n ( s l ) g (20) The time series in Equation (20) is used to predict the future v alue of trust, termed as cumulati v e trust. Definition 5 Cumulative T rust (CT) of a cloud service ( s l ), pr edicted at a time instant n is defined as: C T n ( s l ) = X n i =1 w 0 i C P T i ( s l ) (21) wher e C P T i ( s l ) is a customized pr esent trust of cloud service ( s l ) at time instant i , w 0 i is a weight assigned to it suc h that 0 < w 0 i < 1 and P n i =1 w 0 i = 1 . C P T v alues at latest time instances, which represent recent quality of a cloud service, are more rele v ant in prediction of C T , than the C P T v alues at prior time instances, which represent earlier quality of a cloud service. Hence, e xponentially decreasing weights are assigned to the C P T v alues, starting from the latest C P T v alue to the C P T v alues at prior time instances. This is done using a smoothing f actor such that 0 < < 1 . Thus, the v arious weights assigned to corresponding C P T v alues are: w 0 n = , w 0 n 1 = (1 ) ,. . . , w 0 2 = (1 ) n 2 and w 0 1 = (1 ) n 1 . It is recommended that the v alue of should be set in the range from 0.1 to 0.4. This allo ws the predicted cumulati v e trust to match closely with the computed customized present trust of the service. 5. ALGORITHM FOR ELASTIC TR UST COMPUT A TION Algorithm 1 sho ws the steps for elastic trust computation of a cloud service o v er multiple time windo ws. The algorithm tak es a set of cloud service attrib utes, a number of time instances and the number of time windo ws as input for trust assessment of a cloud service. A set of preferences tak en as another input indicates the requirements of a particular user about the v alues of v arious attrib utes of a cloud service. The algorithm gi v es the output as sets of customized present trust and cumulati v e trust v alues for service s l o v er the time windo ws. The steps of Algorithm 1 for each time windo w , are e xplained as belo w . Step 1. (line 7) The e vidence f actors for the cloud service are acquired and the resultant e vidence matrix C is formed, as sho wn in Equation (1). Step 2. (line 8) The preferences for service attrib utes are combined with the e vidence matrix, to obtain the augmented matrix C P , as indicated in Equation (2). Step 3. (line 9) Normalization function tak es the augmented matrix as input and transforms all the v alues in the matrix to uniform range as specified by Equations (3) to (6). It results into the normalized augmented matrix N C as gi v en by Equation (7). Step 4. (line 10) At this point, the algorithm in v ok es a function to compute weights for v arious attrib utes of the cloud service. The details of the function to compute weights are specified by Algorithm 2 in Section 5.1. Step 5. (line 11) Here, a function is in v ok ed for computation of distances for the v arious attrib utes of the cl oud service, from the specified preferences of attrib utes. The details of the function to compute distances are gi v en by Algorithm 3 in Section 5.2. Step 6. (lines 12 - 17) At each instant of time, computation of customized present trust is performed based on the IJECE V ol. 8, No. 1, February 2018: 304 325 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 313 Algorithm 1 Elastic T rust Computation for cloud service s l 1: Input: a. Set of m cloud service attrib utes, ( AC ) = f R 1 ; R 2 ; :::; R m g b . Number of time instances ( n ) c. User preferences, ( P R ) = f pr 1 ; pr 2 ; :::; pr m g // Preferences for service attrib utes d. Number of time windo ws for trust assessment ( num time windows ) 2: Output: a. Set of Customized Present T rust v alues for service s l , LP = f C P T [1] ; C P T [2] ; :::; C P T [ n ] g b . Set of Cumulati v e T rust v alues for service s l , LC = f C T [1] ; C T [2] ; :::; C T [ num timew indow s ] g 3: Begin 4: step = n ; 5: j = 1 ; 6: while j num timew indow s do 7: Matrix C = Get e vidences( s l ,A C,n) ; 8: Matrix CP = Get augmat(C,A C,PR) ; 9: Matrix NC = Normalize augmat(CP ,A C) ; // Function f N E in Definition 1 10: Set W = Compute weights(NC,A C,n) ; // From Algorithm 2, W is a set of weights of m attrib utes 11: Matrix PN = Compute sumdist(NC,A C,n,W) ; // Matrix PN of summati v e distances computed by Algorithm 3 12: i = 1 ; 13: while i n do 14: Compute Customized Present T rust of service s l at time instant i as: C P T [ i ] = S P i S P i + S N i ; // Function f C P T in Definition 1 and S P i , S N i are elements of matrix PN 15: Add C P T [ i ] in set LP ; 16: i = i + 1 ; 17: end while 18: Compute Cumulati v e T rust of service s l as: C T [ j ] = P n i =1 w 0 i C P T [ i ] ; // Function f C T in Definition 1, w 0 i is a weight assigned to C P T [ i ] 19: Add C T [ j ] in set LC ; 20: PR = Get updatepr ef(A C) ; 21: n = n + step ; 22: j = j + 1 ; 23: end while 24: End summati v e positi v e and ne g ati v e distances of e vidence f actors. The computed v alue is added to the output set of cus- tomized present trust v alues. The details of computation of customized present trust are presented in Section 4.2. Step 7. (lines 18 - 19) Consequently , assessment of cumulati v e trust is performed for the ne xt time instant by using the customized present trust v alues of dif ferent time instances within a time windo w . The computed v alue is added to the output set of cumulati v e trust v alues. The details of computation of cumulati v e trust are elaborated in Section 4.5. Step 8. (lines 20 - 23) The algorithm, in v ok es a function to get the changes in preferences of the particular user , for the attrib utes of a service. The number of time instances for the trust assessment in ne xt time windo w is updated. Accord- ingly , the algorithm continues for the reassessment of the trust of the cloud service o v er subsequent time windo ws. Thus, the algorithm reflects elastic trust computation of a cloud service according to the dynamically changing prefer - ences of the user o v er multiple time windo ws. 5.1. Algorithm f or Computation of W eights Algorithm 2 tak es a normalized augm ented matrix, a set of cloud service attrib utes and a number of time instances as input. The algorithm returns the set of weights for the attrib utes of a cloud service, as the ou t put. As sho wn in the algorithm, the utility de gree for each attrib ute, is computed. From the v alues of utility de gree, weight of each attrib ute is computed. The details of computation of weights are presented in Section 4.3. 5.2. Algorithm f or Computation of Distances fr om Pr efer ences Algorithm 3 tak es a normalized augmented matrix, a set of cloud service attrib utes, a number of time instances and a set of weights for the attrib utes as input. It in turn, gi v es a matrix P N of summati v e positi v e and ne g ati v e Pr efer ences Based Customized T rust Model for Assessment ... (Shilpa Deshpande) Evaluation Warning : The document was created with Spire.PDF for Python.