Service Sharing with Trust in Pervasive Environment: Now it’s Time to Break the Jinx Sheikh Iqbal Ahamed Munirul M. Haque Nilothpal Talukder Department of Math., Stat., and Computer Science Marquette University, Milwaukee, Wisconsin, USA Department of Computer Science Purdue University, West Lafayette, Indiana, USA Department of Math., Stat., and Computer Science Marquette University, Milwaukee, Wisconsin, USA iq@mscs.mu.edu mhaque@cs.purdue.edu ntalukde@mscs.mu.edu ABSTRACT In such as a highly dynamic and open environment, it has become a challenge to deploy multiple context-sensitive services due to the unwillingness of service provider to share resources. This apprehension to share resources stems mainly from a lack of trust. Sometimes infrastructure plays pivotal role to solve this issue with dynamic access control to replace traditional static policies. But when it comes to effective resource sharing in an infrastructure-less environment we face problems such as poor storage and computational capability. In this paper, we have developed a lightweight and distributed trust model based on recommendation, which will guarantee that service providers can securely share an unlimited number of resources, limited only by their hardware and bandwidth limitations. The multi-hop recommendation protocol incorporates a flexible behavioral model to handle interactions during service sharing and usage. This protocol will also assess risk using recommendations from context-sensitive services, in the trust framework, to help ensure smooth access to resources and services. Keywords Access Control, Trust Model, Malicious recommendation, student’s t-distribution. 1. INTRODUCTION The open and dynamic nature of the pervasive computing environment allows entities to join and leave frequently. This causes a large number of autonomous entities to interact in an ad-hoc manner, raising concerns as to the trustworthiness of the service providers. Thus service providers are not willing to share their resources to the anonymous nodes for fear of a potential security violation. To handle the issue of anonymous nodes, a trust management system must grant user access to resources and information based on trustworthiness rather than the conventional technique map access rights [1] authorization. However, in an environment lacking infrastructure, as such an ad-hoc network, it is a challenge to deploy a dynamic access control due to the lack of memory and computational power of the small handheld devices that make up the network. To meet the security goals in a pervasive resource sharing environment, Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. SAC’08, March 16-20, 2008, Fortaleza, Ceará, Brazil. Copyright 2008 ACM 978-1-59593-753-7/08/0003…$5.00. the model has to be able to deal with devices and environment of unknown origin and also has to be adaptive to the dynamics of mobile and socially motivated computing [1]. For this reason, commonly used mechanisms such as role-based access control and public key infrastructures have trouble adapting to these requirements, because they require a predetermined list of entities to properly function. Society, for example, uses trust and reputation to empower a person with pre-established rights. This trust is a dynamic parameter which evolves over time. Our goal is to create a computational model to emulate the complex social mechanism of trust. Several researchers have focused on develop dynamic access control policies, and more specifically on event and location context. Researchers in MIT’s CSAIL lab took a semantic approach in developing a model for context-aware dynamic access control framework [2]. The efforts in [3, 4] mostly emphasized the event and location contexts. The model in [5] uses central authority, which is a system where reliability and accountability are most important, such as healthcare applications. Much research has also been done on trust models in a pervasive computing environment. Several well known trust models depicted in [7, 9, 15, 16, 17] do not adapt well to the pervasive computing environment due to problems including complex computation model (using different cryptographic and secret key sharing techniques) such as [9] (threshold cryptography), [15, 16, 17] (certificate, key binding, and trust relationship). These models utilize protocols to heavy in weight for pervasive devices with limited storage and energy. Due to the presence of fixed infrastructure or central server trust models [15] (a fixed device has to act as security agent in each domain) and [7] (requirement of central key server) are inappropriate for the pervasive environment. A probabilistic approach has been adopted in the Beta Reputation System [34], B-trust [23], and Learning Model [21]. Along with these models, a large number of recent trust models fall into the distributed and dynamic category [25-33]. However, none have explored a protocol to handle malicious recommendations with the exception of the learning model [21]. However, their approach, does not change the overall recommendation value. Our approach addresses the issue of sharing resources in a pervasive environment by incorporating dynamic access control with a trust framework. We focus on developing a recommendation based trust model that will allow the dynamic property of trust to go beyond the traditional methodologies of static role based access control. Our model is also capable of handling malicious recommendations by incorporating a statistical model to help make the access control more reliable and flexible for an open environment. This model will work in an open and dynamic environment where entities frequently join and leave, don’t have any presets access values and don’t require any central authority to issue a certificate. Contributions of this paper: We have summarized the contributions of the paper here: The major contribution is the secured service sharing solution for the open and dynamic pervasive environment. The solution is based on a Hop-based recommendation trust protocol which provides support for the dynamic evolution of trust through a behavioral model. The trust based model also attempts to identify malicious users and weeds out using trust measures. We have also identified a suitable list of index for measuring the users’ willingness to share services in an open environment. The index is used to demonstrate user experience in the evaluation section. The outline of the paper is as follows: The overview of the trust model has been presented in section 2. The summary of malicious recommendation handling is presented in section 3. Related works on trust are further discussed in section 4. Section 5 presents a user evaluation of our system prototype. Our future plans to extend the model are mentioned in section 6. 2. RESOURCE SHARING THROUGH DYNAMIC ACCESS CONTROL WITH TRUST FRAMEWORK 2.1 Why Trust Framework in Access Control Consider a scenario in which node A wants to share or to get access to node B’s resources. The first thing B will do is to reason about the trustworthiness of A. B will accomplish this by analyzing accumulated data from the previous interactions or requesting some recommendations from his trusted parties in the case that A has not had any interactions with B before. There may also be a situation where there might not be enough information to trust, then B has to make his decision based on other variables [1]. Dynamic Access control Shared Services Decision Model Trust Model through a weighted average of direct trust and recommended trust. A decision module is responsible for making a decision based on the trust value and will directly contribute to the policy of the access control. Decision Model Overall Trust Direct Trust Recommended Trust Recommendation Protocol Experience Recommendation Active Passive Discrete Context Fig. 2. Architecture of our Formalized Trust Framework We consider the range of trust value as [0,1]. At first when a node joins a network, it gets an initial trust value of 0.5. Since this node does not have any prior interaction records or known history, they can be neither trusted nor distrusted. 2.2.1 Direct Trust Direct trust evolves from a node’s direct experience with other nodes. As a node interacts with other nodes in the network, its direct trust value for each of the other nodes changes based on the satisfaction level of the interactions. It is the most reliable portion of overall trust. This direct interaction in Figure 3 is shown by a direct link between A and B in the topology of interaction records. 2.2.2 Recommended Trust Recommended trust is used in absence of a direct trust value and is obtained when one node uses suggested trust values from the nodes with which it has direct trust. In figure 3 A might want to have a recommendation from B whether or not to serve C. We devised a general equation for the calculation of recommended trust. Let’s consider a node ωz requests a context ci from ω1. If ω1 (Service Provider, or SP) does not have a direct trust value for ωz (Service Requester, or SR), then it needs to know the recommended trust value to make the context sharing decision. Let us assume that there are n paths (p1, p2, p3,……..,pi,…,pn) with a hop length greater than 1 from ω1 to ωz. Context Service Delivery Fig. 1. Integration of Trust Model into the Dynamic Access Control Framework Because B cannot also allow access to his resources for an indefinite amount of time, his access policies will be dynamically updated on the information based on trust over time. The service delivery agent running on B will consult the access control to decide on access. If trust values are satisfactory A is immediately provided access. The interaction will also be used to modify the existing trust status of A. 2.2 Trust Framework First, we will highlight the points on formation of trust in out model. We define the term “trust value” as the varying degree of trustworthiness. Both types of trust values (direct and recommended) have evolved from pre-evaluated contexts. Each node will have its base level for trust with some other nodes recalculated after each interaction. This base level will directly contribute to the trust values. Overall trust is formed T (1, 2) T ( 2, 3) ... T (x, y ) T (y, z ) ) ( 1) (1 ) .................................(1) 10 Tpi ( Where, 1, 2,...,x, y, z are the nodes on the path pi from SP (ω1) to SR (ωz). D(xT (ci)y), T (x, y) where D(xT (ci)y) xTy, otherwise = Hop distance between ω1 (SP) and ωz (SR) =Distance based aging factor The recommended trust value of ω1 on ωk is calculated as: n R( 1T (ci )z ) Tpi i 1 n ………………………(2) The term (1 ( 1) ) has been used as a weight factor to 10 satisfy the ‘distance based aging’ property. Justification of can be used from Ahamed et al.’s approach [35]. D F C E H L B M A K J N P G I O 10 Fig. 3. A topology of nodes with trust relationships 2.2.3 Active, Passive and Discrete Recommendation Active recommendation is possible only from neighboring nodes, for Passive recommendations the node might consider all path that has hop length >=2. Again when a SP node can’t reach any path to consider it for recommendation, it needs some way to resolve the issue. That’s what we term discrete recommendation. For the same context ci it considers recommendation from other nodes that are same discreet graph relative to SR. In Figure 3, if A needs a recommendation for N, the recommendation values for the paths {M,N},{I,N}, and {P,N} are considered. Here equation (1) takes the following Algorithm for HBRReq: begin for (all neighboring nodes of SP) UNICAST(HBRRequest(Req_ID,SP_ID, SR_ID,IH,IR,TS)) while(until response from all nodes or timeout ) begin R= Receive(HBRRep(Req_ID,Rec_ID,RH,TR,TS)); Rec_Wo_Aging = (TR/(IH-RH)) ; if (Rec_Wo_Aging >=0.5) begin Rec_W_Aging = Rec_Wo_Aging * (1 ((IH RH ) 1) ) form T (i, z ) Tp i ( ) T (i, z ) [ 1, Considerin g only 1 hop paths] Since we are getting recommendations from nodes that are in no way connected to SP, we used =0.5 which is a relatively lower weight factor. A later equation (2) is used to find the discrete recommendation value. 2.3 Hop Based Recommendation Protocol (HBRP) We are using a hop based recommendation protocol to determine trust values to consider a node eligible for access. Here we used the following notations: SP = Service Provider, SR = Service Requester, Req_ID = Request ID, Rec_ID = Recommender ID, IR = Initial Recommendation, TR = Total Recommendation, IH = Initial Hop value, RH=Reduced Hop value, TS = Time stamp. This protocol actually includes mechanisms for active and passive recommendations. The hop field (IH) defines the maximum path length which enables a node to avoid a long chain of Table 1: Upper-lower limits for different confidence values Confidence LCL UCL Standard Mean Level (%) Deviation 90 0.645381 0.886619 95 0.766 0.61715 0.91485 0.2082 99 0.55216 0.97984 recommendations. This value is reduced in each hop by 1 and the path is ignored when the field becomes 0. The IR field contains the trust value of the first link over the path. The TS field has been used to restrict a replay attack. The RH field shows the hop value which has been formed by reducing the IH value by one in each hop. The TR field sums up the trust value over the path. if (Rec_W_Aging > 0.5) begin Rec = Rec + Rec_W_Aging Total_No_Rec = Total_No_Rec + 1; end else Rec_W_Aging = Rec_Wo_Aging (1 (( IH RH ) 1) ) 10 * Rec = Rec + Rec_W_Aging Total_No_Rec = Total_No_Rec + 1; end if (Total_No_Rec == 0) Rec_Trust = 0; else Rec_Trust = Rec/Total_No_Rec; return Rec_Trust; end We can use the recommendation protocol from Ahamed et al.’s [35] approach in its discreet variant and also for the behavioral model that updates trust value upon each interaction. 3. RISKS IN TRUST FRAMEWORK: HANDLING MALICIOUS RECOMMENDATIONS Sometimes a node is in a scenario where the recommendation value contrasts the current recommendation value. We term these values malicious recommendations, which can impact the overall value of the node. There can be two such situations. a) When a malicious node gives a high recommendation value for a node when the overall value is poor. b) When a malicious node recommends a very low value contrasting high recommendations from others. We have adopted a statistical method to address this issue of malicious recommendation. Our assumption is that the number of benevolent nodes is much larger compared to the number of malicious nodes. This protocol is based on a statistical distribution named Student’s t-Distribution [36, 37]. This approach is used when the standard deviation is not known which is true for almost every real life scenario. For example, let us take a data set of 10 recommendations which are as follows: {0.9, 0.88, 0.85, 0.75, 0.78, 0.8, 0.73, 0.2, 0.87, 0.9}. In this distribution a confidence value of 90% indicates that the mean of the data set will be in the interval specified by the UCL (Upper Confidence Limit) and LCL (Lower Confidence Limit). We have used the method t-Estimate : Mean for finding the following results for different confidence levels. As we can see from the table, the interval region defined by UCL and LCL varies inversely with the confidence level. We do not want to leave out any recommendation value unless it is an obvious outlier. We have used a confidence level of 99% to achieve quite a large interval region. Table 2: Comparison of various trust based model Model Infrastructure Support Needed Adaptiv e Lightweight Dynami c Distri buted Context specific Behavioral model Risk model DB A TBA B-trust [13] N/A N Y Y Y N/A Y N N Y SSRD [25] Y N Y Y Y Y N N N N PTM [26] N/A N Y Y Y N L N N Y FDTM [27] Y Y N Y Y N N N N N LMT [11] Y N Y Y Y N L N N Y TUTC [28] Y N Y Y Y N N Y N Y TMUS [12] Y N Y Y Y N L N N Y TRAC [30] N N/A N Y Y N N N N N/A TPAN [24] Y N Y Y Y N N N N N/A TFPC [31] N/A N Y Y Y N N N N N CBTM [32] N/A N L Y Y N N N N N MobileTrust [14] Trust based Service sharing model Y N Y Y Y Y L N N Y Y Y Y Y Y Y Y N Y Y where, α = number of times a device sends a recommendation which falls in the interval (+1σ, +2σ] or (-1σ, -2σ] β = number of times a device sends a recommendations which falls in the interval (+2σ, UNPL] or (-2σ, LNPL] δ = number of times a device sends a recommendation which has a value greater than or equal to UNPL or lower than or equal to LNPL. w1, w2, and w3 are weight factors with w1< w2<w3 M ean LCL 1.2 1 0.8 0.6 0.4 0.2 In the second approach, a device will be discarded if, η >= Δ, where, η = number of times a device sends recommendation which falls outside of the interval defines by UCL and LCL. Rec10 Rec9 Rec8 Rec7 Rec6 Rec5 Rec4 Rec3 Rec2 0 Rec1 Recommendation Value Rec_Val UCL Recom m endation Fig. 4. Plot result of the recommendation values using tEstimate : Mean We have used the information of interval region for two reasons: 1. We discarded any recommendation value which is outside this interval and took the mean of the rest of the values as the overall recommendation value. The optimization is readily observable. The old overall recommendation value of 0.766 has been replaced by a much accurate overall recommendation value of 0.828889 ((0.9+ 0.88+ 0.85+ 0.75+ 0.78+ 0.8+ 0.73+ 0.87+ 0.9)/9). 2. If the recommendation value given by a specific device falls outside the interval region for more than a specified threshold value (Δ), the device will be discarded. In the first approach, a device will be discarded if, w1α + w2β + w 3δ ………………………………………………........(3) >= Δ The value of Δ is dependant on the level of importance of security the environment. In a highly sophisticated environment a very small number of out of range recommendations will put a device out of a network, whereas more tolerance can be shown in a less sophisticated environment. 4. RELATED WORK There has been a good number of works on focusing on dynamic access control for resource sharing and interaction, as we have listed in this section [11-14, 25-32]. A comparison table of trust models is presented in Table 2. The following criteria are focused on: Risk Model Distance Based Aging(DBA) Time based Aging(TBA) Lightweight Dynamic Distributed Context specific Behavioral model Adaptability Support of Infrastructure While all of the models support dynamic updates of trust values, they lack other features necessary for the dynamic pervasive environment. Although the more recent works consider time based aging in trust evolution, hop based aging is unique to our model. Other models adopt complex cryptographic schemes which are not suitable for the light-weight models. Most of them lack a behavioral model which is most important for the evolution of trust in such a dynamic model. 5. EVALUATION In this section, we demonstrate two different the evaluation criteria for our service sharing model. First we present the implementation of the secured service sharing application which takes advantage of the trust based access control model. Second, we present a usability survey used to evaluate our applixation. 5.1 Implementation We have implemented a prototype of the secured service sharing application with the Hop-based Trust protocol using VC#.Net Compact Framework. Screenshots of the application are provided in figure 5. The aforementioned application has the ability to discover neighboring service providers and provide a list of services available from these providers. The following screenshots demonstrates a successful and a failed attempt to access shared services. How much will Context specific profiles be used? For example, restriction to time based or location based profile while granting the service access. The more roles, the more the discomfort users experience. So, we provide lowest rating for the highest number of profiles. (Context profiles) To what extent the user roles will be distinguished? For uniformity of the discomfort level, the lowest rating goes to highest number of roles. (Role extent). Each of these characteristics was rated on scale from 1 through 5, 5 being the highest rating. Low ratings imply the user is unwilling to share the services. The services taken into consideration for the survey were: Date/Time Weather WordPad Chat Unzip SW Address book These usability survey results show that a service’s trust based access controls had a positive impact on the users’ willingness to share services. After being provided with the service sharing application, the users’ willingness to share shifted towards positive outlook. Although sensitive services like address book don’t show high rated shift, still the change is still visible. Usability (Before) 5 4.5 4 Frequency 3.5 3 Access 2.5 2 Context Profile Role Extent 1.5 1 0.5 0 Date/Time Weather Wordpad Chat softw are Unzip SW Address book (a) Usability (After) 5 4.5 Fig. 5. Screenshots of the Service sharing application based on trust based access control 5.2 Usability survey We have conducted a usability survey among 20 subjects to examine the user experience based on services offered through our trust based access control model. The survey was conducted in two phases. The first phase offered the users a mock up of screens to allow the users to familiarize themselves with the handheld device. In the second phase, the user was provided with the actual services in order to see if their expectations were met or exceeded. The difference shows a positive shift in the user’s expectations (Fig. 6). The following characteristics have been identified based on the usability survey. How frequently the services will be shared? (Frequency) How many accesses per service will be provided by default? User might consider affect of bandwidth and quality. (Access) 4 3.5 Frequency 3 2.5 2 Access Context Profile 1.5 1 Role Extent 0.5 0 Date/Time Weather Wordpad Chat softw are Unzip SW Address book (b) Fig. 6. Usability survey (a) before and (b) after the service usage 6. CONCLUSION AND FUTURE WORKS In this paper, we presented a trust model to fit the dynamic access control framework intended for pervasive environment. We used this information to optimize the accuracy of the recommendation process and the discarding of malicious devices from the network. The prototype of the secured service sharing application presented in the evaluation section uses this hop-based recommendation protocol. We are integrating the behavioral model and the dynamic update of trust with a prioritized list of services into the application. We are also incorporating the risks involved in the different sharing scenarios and working towards a complete risk model. As a continuous addendum to the features, this access control module will be placed in the MARKS ((Middleware Adaptability for resource Discovery, Knowledge Usability, and Self Healing) [38] middleware. Apart from security issues in service sharing, our future research lies with privacy issues that may arise due to context-awareness of applications in the pervasive environment [39]. 7. REFERENCES [1] Philip Robinson, Harald Vogt, Waleed Wagealla: “Some Research Challenges in Pervasive Computing,” Book Chapter, p. 1-16, Privacy, Security and Trust within the Context of Pervasive Computing, Springer, 2004. [2] Alessandra Toninelli, Rebecca Montanari, Lalana Kagal, and Ora Lassila: “A Semantic Context-Aware Access Control Framework for Secure Collaborations in Pervasive Computing Environments”, 5th International Semantic Web Conference, 2006. [3] Covington, M.J., et al.: “Securing Context-Aware Applications Using Environmental Roles”, In: Proc. of the 6th ACM Symposium on Access Control Models and Technologies (SACMAT 2001), May 3-4, Chantilly, Virginia, USA ACM (2001). [4] Nicholaos Michalakis, “Location Aware Access Control for Pervasive Computing Environment”, MIT Lab of Computer Science. [5] M. Anisetti, C.A. Ardagna, V. Bellandi, E. Damiani, S. De Capitani di Vimercati, P. Samarati , “OpenAmbient: a Pervasive Access Control Architecture”, Co-located with the International Conference on Emerging Trends in Information and Communication Security (ETRICS’06), Freiburg, Germany, June 6-9, 2006. [6] Kui Ren and Wenjing Lou: “Privacy Enhanced Access Control in Pervasive Computing Environments” [7]Garfinkel, S. (1995): PGP : Pretty Good Privacy, O’Reilly & Associates, Inc. [8] A. Abdul-Rahman and S. Hailes, “A Distributed Trust Model,” Proceedings of the 1997 workshop on New security paradigms, 1998. [9] Zhou, L. and Haas, Z. J. (1999): Securing Ad Hoc Networks, IEEE Network Magazine, 13(6). [10] H. Sun and J. Song, “Strategyproof Trust Management in Wireless Ad Hoc Network”, Proceedings of the IEEE Canadian Conference on Computer and Electrical Engineering, 2004. [11] Markus C. Huebscher, Julie A. McCann, "A Learning Model for Trustworthiness of Context-Awareness Services," percomw, pp. 120-124, Third IEEE International Conference on Pervasive Computing and Communications Workshops (PERCOMW'05), 2005. [12] Hassan Jameel, Le Xuan Hung, Umar Kalim, Ali Sajjad, Sungyoung Lee, and Young-Koo Lee, “A trust model for ubiquitous systems based on vectors of trust values,” In Seventh IEEE International Symposium on Multimedia, 12-14 Dec. 2005 Page(s):6 pp. [13] D. Quercia, S. Hailes, and L. Capra, “B-trust: Bayesian Trust Framework for Pervasive Computing,” iTrust. LNCS. May 2006. Pisa, Italy. [14] Ching Lin and Vijay Varadharajan, “Trust enhanced security for mobile agents”, In The Seventh IEEE International Conference on E-Commerce Technology CEC 2005, Page(s):231 - 238 [15] A. Josang. The right type of trust for distributed systems. In New Security Paradigms´96 Workshop, 96. [16] A. Josang. An algebra for assessing trust in certification chains. In Proceedings of the Network and Distributed Systems Security (NDSS’99), 99. [17] A. Josang and S. J. Knapskog. A metric for trusted systems. In Proceedings 21st NIST-NCSC National Information Systems Security Conference, pages 16–29, 98. [18] L. Kagal, T. Finin, and A. Joshi. Trust-based security in pervasive computing environments. In IEEE Computer, volume 34, pages pp. 154–157, December 2001 [19] A. Abdul-Rahman and S. Hailes. Supporting trust in virtual communities. In Proceedings 33th Hawaii International Conference on System Sciences. IEEE Press, January 2000. [20] Stajano, F. and Anderson, R. (1999): The Resurrecting Duckling: Security Issues for Ad hoc Wireless Networks, Proc. of the 7th International Workshop on Security Protocols, 1796:172-194. [21] R. Guha, R. Kumar. Propagation of trust and distrust. In WWW’04: Proceedings of the 13th international conference on World Wide Web, pages 403–412. ACM Press, 2004. [22] M. Kamvar, S.D. Schlosser. The eigentrust algorithm for reputation management in p2p networks. In WWW’03: Proceedings of the 12th international conference on World Wide Web, pages 640–651. ACM Press, 2003. [23] R. Matthew, R. Agrawal. Trust management for the semantic web, 2003. [24] J. Theodorakopoulos, G. Baras. Trust evaluation in ad-hoc networks. In WiSe ’04: Proceedings of the 2004 ACM workshop on Wireless security, pages 1–10. ACM Press, 2004. [25] M. Sharmin, S. Ahmed, and S. I. Ahamed, “An Adaptive Lightweight Trust Reliant Secure Resource Discovery for Pervasive Computing Environments,” Proceedings of the Fourth Annual IEEE International Conference on Pervasive Computer and Communications (PerCom2006), Pisa, Italy, Mar 2006, pp. 258-263. [26] F. Almenarez, A. Marin, C. Campo, and C. Garcia, “PTM: A Pervasive Trust Management Model for dynamic open environments,” Pervasive Security, Privacy, and Trust (pspt 2004), Massachusetts, 2004. [27] Daoxi Xiu and Zhaoyu Liu, “A Dynamic Trust Model for Pervasive Computing Environments”, The Fourth Annual Security Conference, Las Vegas, NV, March 30-31, 2005. [28] Brian Shand, Nathan Dimmock, Jean Bacon, "Trust for Ubiquitous, Transparent Collaboration," percom, p. 153-160, First IEEE International Conference on Pervasive Computing and Communications (PerCom'03), 2003. [29] Hassan Jameel, Le Xuan Hung, Umar Kalim, Ali Sajjad, Sungyoung Lee, and Young-Koo Lee, “A trust model for ubiquitous systems based on vectors of trust values,” In Seventh IEEE International Symposium on Multimedia, 12-14 Dec. 2005 Page(s):6 pp. [30] J. Basu and V. Callaghan, “Towards a Trust Based Approach to Security and User Confidence in Pervasive Computing Systems,” the IEE International Workshop, Intelligent Environments 2005 (IE05), UK, June 2005, Accessed May 2006, from http://cswww.essex.ac.uk/ Research/iieg/papers/Jisnu%20Basu%20IE05.pdf [31] A. Pirzada and C. McDonald, “Establishing Trust in Pure Ad-hoc Networks” in Proceedings of the 27th conference on Australasian computer science, Vol 26, 2004 [32] Steven T. Wolfe, Sheikh I. Ahamed, and Mohammad Zulkernine, “A Trust Framework for Pervasive Computing Environments”, The 4th ACS/IEEE International Conference on Computer Systems and Applications (AICCSA-06), IEEE CS Press, Dubai, UAE, March 2006, pp. 312-319. [33] Rui He, Jianwei Niu, Man Yuan, Jianping Hu, "A Novel Cloud-Based Trust Model for Pervasive Computing," cit, pp. 693-700, The Fourth International Conference on Computer and Information Technology (CIT'04), 2004. [34] R. Ismail. The beta reputation system. In Proceedings of the 15th Bled Conference on Electronic Commerce, 2002. [35] Munirul Haque, Sheikh I Ahamed, “An Omnipresent Formal Trust Model (FTM) for Pervasive Computing Environment”, Proceedings of the 31st Annual International Computer Software and Applications Conference (COMPSAC 2007), IEEE CS Press, Beijing, China, July 23-27, pp. 49-56. [36] Student" (W.S. Gosset) (1908) The probable error of a mean. Biometrika 6(1):1—25 [37] M. Abramowitz and I. A. Stegun, eds. (1972) Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables. New York: Dover. (See Section 26.7 [38] M. Sharmin, S. Ahmed, and S. I. Ahamed, “MARKS (Middleware Adaptability for Resource Discovery, Knowledge Usability, and Self Healing) in Pervasive Computing Environments,” Proceedings of the Third International Conference on Information Technology: New Generations, USA, Apr 2006, pp. 306-313. [39] S. I. Ahamed, N. Talukder, M. Haque, "Privacy Challenges in Context Sensitive Access Control for Pervasive Environment", in the First International Workshop on the Security and Privacy of the Ubiquitous Communications Systems (SPEUCS) in conjunction with Mobiquitous 2007.