Study On Trust Evaluation Method Based On Intuitionistic Fuzzy Theory

With the rapid development and prevalence of open network applications,such as peer-to-peer networks,e-business,social networks and mobile commerce,etc,open network systems have become the important tools and platforms for people's interactions,including sharing resource,shopping,social activities,and so on.However,duo to the open network environment has the characters of fuzziness,randomness and uncertainty,customers confront lots of secure issues when selecting totally unknown object entities.Under these uncertainty environments,the traditional secure mechanisms depend mainly on the CA(certificate authority)and PKI(public key infrastructure)which is difficult to adapt to these requirements.Duo to anonymity and autonomy of object entities,every evaluation entity has to evaluate trustworthiness of other entities and select the most trusted one.Consequently,it is necessary to design an effective trust mechanism to help user select appropriate object,improve the success rate of cooperation,evaluate unknown entities and ensure smooth running of the whole system.Existing trust models in open distributed systems(e.g.,P2P network,e-business,social network,or mobile commerce)still have following weaknesses:(1)Most trust models cannot well address the ambiguity and uncertainty of trust;(2)Unfair recommendations,denigrate or exaggerate behaviors can be provided by malicious entities;(3)Trust transitivity usually overlooks trust quality of path;(4)Existing recommended trust evaluation models cannot effectively deal with heterogeneous information.To address above issues,intuitionistic fuzzy theory(IFT)is introduced into the study of trust model.As an extension of fuzzy set,intuitionistic fuzzy(IF)set(IFS)has better agility in expressing the uncertainty and ambiguity.This dissertation researched trust evaluation methods based on IF information,including trust representation,generation,transitivity,decision,malicious behavior resistance and heterogeneous recommendation information processing.The contributions of this dissertation are fourfold:(1)Existing trust models cannot effectively express the uncertainty of trust relationship and deal with such issues as dishonest feedbacks and strategic frauds from malicious entities in the distributed network.To address those weaknesses,an adaptive trust model based on IF information(IFI)is proposed.There are three key issues being addressed in this model.The first one is to construct a method on aggregating IFI to compute the direct trust which contains latest permanence factor and time decay factor.The second one is to detect dishonest recommendation using the indexes of recommendation credibility and uniformity.The third one is to develop an adaptive method to determine the weights of direct and indirect trust automatically.The simulation experiments demonstrate that the proposed model is not only robust on malicious attacks,but also has better adaptability and effectiveness.(2)In the e-commerce environment,There is complexity,fuzziness and uncertainty and so on,which can't ensure consumer pay.We urgently need to judge service provider's credibility and ability to provide services.Trust provides guarantee for quality of service.Taking the e-commerce environment as the research background,by use of IF theory,we proposed a trust model based on feedback information.This trust evaluation approach can implement the conversion between feedback attribute and IF number(IFN).The weight of attributes was obtained by utilizing entropy for IFSs.Then,intuitionistic weighted arithmetic averaging operator was used to aggregate the IF information corresponding to each entity,which comprehensively,objectively,delicately reflected the fuzziness and uncertainty of the trust.Based on the above method,we develop an adaptive aggregation method and a new trust score considering risk preference.And a subjective trust model based on consumers' risk preference was proposed.The example and simulation results show the feasibility of the trust evaluation approach.A sensitivity analysis for risk appetite shows that different service requestor risk appetite will lead to different service provider's trust score.It also verified that the approach can effectively inhibit attack from malicious node.The trust evaluation approach in this paper may provide a new way for research of trust evaluation in e-commence environment.(3)Aiming at fuzziness and subjective problems that exist in trust evaluation for social networks,and it is very difficult to quantify and forecast trust degree of the unfamiliar entities,IFS is introduced into trust evaluation which can represent trust degree,distrust degree and hesitancy degree;a new multi-dimensional trust transitivity model based on intuitionistic fuzzy theory(IFT)is proposed.First,the method discusses IFS-based trust propagation operator and its rationality.Next,through integrated into account the influence of the length and trust quality of paths,the paper studies two weighted trust aggregation operators.Finally,a social network analysis(SNA)methodology to represent and model trust relationship between entities and to compute the trust degree of each path is developed.It can be seen from the simulation results which compared with the existing trust transitivity models,the proposed model has high prediction accuracy.(4)In actuality evaluation of mobile technology,the ratings under the attributes are in the form of a variety of multiple data types,the familiarity with the evaluation areas(called confidence levels)of recommenders are different.However,existing recommended trust evaluation models cannot be used to these problems described above effectively.In this dissertation,a heterogeneous recommended trust evaluation models based on intuitionistic fuzzy information are presented.In this model,the conversion between heterogeneous evaluation information and intuitionistic fuzzy information is implemented for describing the subjective and fuzziness of the ratings.To overcome the double uncertainty from the subjective judgment of the weights of recommenders,an entropy method is developed based on the original evaluation information.Considering the confidence levels of recommenders,we develop a confidence interval-value IF weighted averaging(IIFWA)operator.Finally,an example about trust evaluation of suppliers in mobile commerce is given to verify the practicality and effectiveness of the proposed models.