基于直觉模糊理论的信任评估方法研究

发布时间：2018-03-03 01:14

本文选题：信任模型　切入点：直觉模糊集　出处：《江西财经大学》2015年博士论文　论文类型：学位论文

【摘要】：随着P2P网络、电子商务、社会网络、移动商务等大规模开放式应用系统的普及和发展,开放式网络环境已成为人们分享资源、购物、社交活动、服务推荐等在线交互活动的重要平台和工具。然而,开放式网络环境具有模糊性、随机性和不确定性的特点,使用户在选择目标实体交互时面临了诸多风险。在这种不确定的网络环境下,传统的基于CA,PKI等技术的静态安全机制难以适应这种环境的需求。由于网络实体的匿名性和自治性,每个评估实体必须评估其它目标实体的可信性,并且选择最可信的实体进行交互。因此,设计有效的信任评估机制,在指导用户选择合适的交互对象、提高合作成功率、评估陌生实体、保障开放式计算系统的顺利运行等方面,具有重要的现实意义。然而,现有开放式网络环境下的多数信任模型尚未完善,主要存在以下四个问题:(1)不能很好刻画信任的模糊性和不确定性特点;(2)对不诚实推荐、夸大、诋毁等恶意行为的免疫力不够;(3)信任传递经常忽视路径信任质量的影响;(4)不能有效处理异质推荐信息。为了解决上述问题,论文将直觉模糊理论引入到信任建模的研究中来。作为传统模糊集的扩展,直觉模糊集能够更加细腻地刻画信任的模糊性和不确定性。本文研究基于直觉模糊理论的信任评估方法,包括基于直觉模糊信息的信任表示、获取、分配、传递、决策、恶意行为抵制和异质信息处理等内容,具有重要的理论和实践意义。具体工作如下:(1)提出了一种集成直觉模糊信息的激励自适应信任模型。为了激励提供可信服务的实体,惩罚不良行为实体,该模型结合服务持续性强度和时间衰减因子,给出了实体直接信任直觉模糊数的计算方法;同时利用实体的推荐可信度和行为一致性过滤不诚实推荐行为。除此之外,为了克服综合信任聚合计算中权重确定较主观的问题,给出了一种新的自适应权重分配方法。仿真实验表明,该模型在提高实体服务积极性和抑制恶意欺诈行为中,都有较好的适应性、灵敏性和有效性。(2)提出了一种基于直觉模糊信息的多维反馈信任模型。该模型以电子商务环境为背景,给出了定量和定性两类信任反馈信息集成为直觉模糊数的方法。通过直觉模糊熵确定属性权重,并利用直觉加权算数平均算子求出实体的综合直觉模糊数,从而全面地、客观地反映了信任的模糊性和不确定性。在上述方法的基础上,引入风险偏好因子,研究了消费者风险偏好对信任评估的影响。实例和仿真实验表明,基于直觉模糊信息的多维反馈信任评估方法是可行的;通过对风险偏好因子的灵敏度分析,发现服务请求者的风险偏好不同,则评估的服务提供商的信任得分也不同;同时也验证了该方法可以有效地抑制恶意节点的攻击,为电子商务环境下信任评价的研究提供了新思路。(3)设计了一种基于直觉模糊理论的多元信任传递模型。为了解决复杂网络中陌生实体信任评估问题,同时避免信任传递中信任不确定性的丢失问题,该模型讨论了直觉模糊环境下基本的信任传播算子和聚合算子;经过综合考虑路径长度和信任质量对信任聚合的影响,研究了两种加权信任聚合算子,给出了基于社会网络分析的路径信任质量计算方法。仿真实验表明,与已有模型相比,路径长度和信任质量混合加权聚合策略在信任度量准确性方面有较大提高。(4)构建了一种基于区间直觉模糊信息的异质推荐信任模型。该模型采用区间直觉模糊数表示信任,针对推荐信息的异构性,给出了基于异质信息的信任区间直觉模糊数的获取方法,从而较全面地刻画了评价信息的主观性和模糊性。为了避免推荐者主观赋权带来的二次不确定性,给出了基于原始评价信息的熵权法。考虑到推荐者知识背景的差异性,提出了信心区间直觉模糊加权平均集成算子及其定理和性质,并利用基于信心的集成算子计算综合信任直觉模糊信息。最后,通过一个移动商务供应商的信任评估实例验证该模型的有效性和合理性。
[Abstract]:With the P2P network, e-commerce, social network, the development and popularization of mobile commerce and other large-scale open application system, open network environment has become the people to share resources, shopping, social activities, an important platform and tool service recommendation online interactive activities. However, the open network environment is fuzzy, random and not the uncertainty characteristics, so that users face a lot of risk in the selection of the target entity interactions. In this uncertain network environment, CA based on the traditional static security mechanism of PKI technology cannot meet the environmental requirements. Because the network entity anonymity and autonomy, the credibility of each evaluation entity must evaluate other target entities the interact and choose the most trusted entities. Therefore, the trust evaluation mechanism is effective in design, guide the users to choose the appropriate interactive object, improve the success rate of cooperation, Evaluation of strange entities, open computing system guarantee smooth operation, has important practical significance. However, most of the existing trust model in open network environment is not perfect, there are mainly four problems: (1) can not describe the characteristic of fuzziness and uncertainty of trust; (2) to exaggerate dishonest recommendation, slander and other malicious acts of immunity is not enough; (3) effects of trust transfer often ignore path trust quality; (4) can not effectively deal with heterogeneous information recommendation. In order to solve the above problems, the intuitionistic fuzzy theory is introduced to study the trust modeling in the past. As the expansion of traditional fuzzy sets, fuzzy of intuitionistic fuzzy sets can be more delicate portrait of trust and uncertainty. The trust evaluation method based on intuitionistic fuzzy theory, including intuitionistic fuzzy information representation based on trust, acquisition, distribution, transmission, The decision, malicious boycott and heterogeneous information processing, has important theoretical and practical significance. The specific work is as follows: (1) the paper puts forward an integrated intuitionistic fuzzy information incentive model adaptive trust. In order to provide credible service incentive entities, punish bad behavior entity, the model combined with the attenuation factor of strength and continuous service time gives the entity direct trust calculating method of fuzzy number intuitionistic; while using recommendation credibility and behavior consistency filtering entity dishonest recommendation behavior. In addition, in order to overcome the comprehensive trust aggregation calculation is a subjective matter in determining the weights, presents a new method of adaptive weight distribution. Simulation results show that the the model in improving the entity service enthusiasm and inhibit malicious fraud, have better adaptability, flexibility and effectiveness. (2) proposed a model based on intuition Multidimensional fuzzy information feedback trust model. The model is based on the background of e-commerce, information integration method for intuitionistic fuzzy numbers is given the quantitative and qualitative analysis of two types of trust. Through the feedback of intuitionistic fuzzy entropy to determine attribute weights, and calculated the comprehensive entity using intuitionistic fuzzy number intuitionistic weighted arithmetic average operator, thus comprehensively, objectively reflects the fuzziness of trust and uncertainty. Based on the above method, the introduction of risk factor, study the influence consumers' risk preference on trust evaluation. Examples and simulation experiments, multidimensional intuitionistic fuzzy information feedback based on trust evaluation method is feasible; based on the sensitivity analysis of risk factor. Find the service requester different risk preferences, trust score evaluation service providers are different; the results also show that the method can effectively suppress the malicious node The point of attack, and provides a new idea for the evaluation of trust in e-commerce environment. (3) the design of a multi trust transfer model based on intuitionistic fuzzy theory. In order to solve the complex network strange entity trust evaluation problem, while avoiding the trust transfer of trust and uncertainty of the loss problem, the model of trust is discussed the basic propagator and aggregation operators under intuitionistic fuzzy environment; after considering the path length and affect the quality of the trust trust aggregation, study two kinds of weighted trust aggregation operator is given to calculate the path of trust quality based on social network analysis method. Simulation results show that compared with the existing models, improved path length and trust the quality of the mixed weighted aggregation strategy measure accuracy in terms of trust. (4) to construct a based on interval valued intuitionistic fuzzy information recommendation trust model. The model of heterogeneity The interval intuitionistic fuzzy numbers to recommend trust, heterogeneous information, method for the trust interval intuitionistic fuzzy numbers based on heterogeneous information is given, so as to comprehensively characterize the evaluation information of subjectivity and fuzziness. In order to avoid the presenter at the two subjective uncertainty, given the entropy method to evaluate the original based on the information. Considering the differences of the recommended knowledge background, the confidence interval intuitionistic fuzzy weighted average operator and its integration theorems and properties, and the use of integrated operator confidence calculation based on intuitionistic fuzzy comprehensive trust information. Finally, through the trust evaluation instance of a mobile commerce provider verified the validity and rationality of the model.

【学位授予单位】：江西财经大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TP393.08

【参考文献】