【作者】 王江涛；

【导师】 陈志刚；

【作者基本信息】 中南大学 ， 计算机软件与理论， 2013， 博士

【摘要】 普适计算的网络环境具有动态和不可预知的特点,交互可能发生在熟知的实体之间也可能发生在完全陌生的实体之间。普适计算倡导以人为中心,因此如同人类社会网络的存在的安全与信任问题一样,普适计算技术的发展必然也伴随着安全与信任问题,然而传统的安全技术PKI, CA等静态信任安全机制已经不能适应这种动态变化环境的需求,如果缺乏有效的安全信任机制保障,普适计算将难以真正实现。因此如何设计有效的动态信任机制来适应普适计算中的各种动态变化的环境已经逐渐成为研究的一个热点。但是目前为止所提出的一些普适计算网络环境下的信任模型尚未完善,主要存在以下几个问题,一、绝大多数信任模型,缺乏上下文感知能力,缺乏足够的能够对时间,地点,交互内容等上下文感知的信任计算模型。二、大部分信任模型在计算信任度的时候没有考虑信任关系之间的语义关系,也未能考虑不同实体之间天然存在的语义关系,使得这些模型的效率和精细度都不够。三、大部分信任模型设计的比较复杂,不合适资源受限的普适计算环境。本文针对上述问题进行了深入研究并提出了新的信任模型以部分地解决这些问题,在此基础上,本文对信任模型在普适计算环境中的WSN和P2P网络下的一些应用做了研究,在普适计算环境中信任模型的应用主要集中在两个方面。第一个方面是是用于促进实体间的信任水平进而提高实体间的合作水平,本文在这个方面的研究分别在第五章和第六章针对不同的应用环境提出了两种方法来促进实体间的合作水平。第二个方面的应用是为普适计算条件下的一些具体应用提供安全保障机制,本文第七章在这方面的研究针对具体的WSN网络路由应用提出了一种基于可信路由树的数据路由机制。主要工作如下：(1)针对普适计算网络环境的特点,提出了一种适合于普适计算网络环境的上下文感知的自适应信任模型CASATM,该信任模型能够自适应的对不同的服务提供不同层次的安全保障,能有效地对时间,地点,服务内容等上下文感知。同时提出了一种简单高效的风险评估模型来完成对一些陌生实体的信任度初始化。通过仿真实验分析,该信任模型能有效的感知上下文并抵抗恶意客体的周期性欺骗行为。(2)借助本体论提出了一种基于语义距离的信任模型。该信任模型利用实体之间的语义距离和信任关系之间的语义距离来更加精确的计算信任度。同时在本文的信任模型中区分了实体的行为信任和能力信任,所有的实体在基于自身利益最大化的思想下利用该信任模型来做出决策。仿真实验证实了该模型的有效性,能有效提高普适计算网络环境中实体间交互的成功率与交互效率。(3)对分布式信任模型在普适计算环境下对于提高节点合作水平的应用进行了研究。在分布式信任模型的基础上借助博弈论理论研究了激励节点合作行为的机制,让节点迅速学习适宜的合作行为,与传统的分布式模型不同的是适应度高的行为策略的传播效率更高,这样可以加速整体合作水平提高的收敛过程。与此同时扩展了收益函数,使得实体在低能量状态下可以采用相对保守的行为从而延长整个网络的寿命。(4)在普适计算环境的P2P网络下,研究了信任模型在多agent系统合作模型中的应用。提出了一种基于兴趣信任的多agent合作模型,利用Bayesian方法对agent的信任度按兴趣领域进行了分类评估,通过agent间基于兴趣信任的链接更新形成稳定的多agent合作系统。最后的仿真实验显示,本文提出的基于兴趣信任的合作系统演化方案不仅能有效提高任务的执行成功率,而且能有效改善网络拓扑结构,使所有的agent在不同的兴趣领域内聚集在该领域内服务能力较强的信任agent周围,保证了合作对象选取的可靠性。经过多轮演化后,本方案能使agent在不同的兴趣领域形成“兴趣分类聚集”,减小了网络同化现象,突出了agent的个性。(5)在普适计算环境的无线传感器网络下,研究了信任模型在数据路由中的应用,提出了一种基于信任核心树的数据路由算法。该算法在局部范围内选取具有相对较高剩余能量和信任度的节点作为簇首节点,从而对整个网络进行分簇。然后通过一个最小能耗生成树算法将所有簇首节点组织成以Sink节点为根的信任核心树,进而对该树进行扩展构建成覆盖全网的信任路由树。所有的数据沿着信任路由树上唯一的路径向Sink节点汇报。同时该算法引入了信任模型,该模型通过对网络中有数据包篡改,丢包,谎报等行为的恶意节点进行检测来评估节点的信任度。仿真结果证明该TCTR算法能有效减小能耗、平衡负载从而延长网络生存时间,同时也能有效识别与隔离恶意节点从而提高路径安全度。更多还原

【Abstract】 The network environment for pervasive computing has the characters of dynamic and unpredictable, interactions could be launched between familiar entities or totally unknown entities. As the security and trust problem existed in human society, these problems also exist in the development of pervasive computing technology, however, traditional static security mechanisms like PKI and CA are not appropriate for such open and dynamic environments. Pervasive computing can not be realized without the quarantee of efficient security and trust mechanism. A hot spot research in recent years is how to design an effective dynamic trust mechanism to deal with the dynamic and unpredictable pervasive environments. However, most trust models proposed for pervasive environments is not yet perfect, some disadvantages are. existed as described following:1:Most trust models lack of the ability of context awareness and adaptiveness, contexts like time,location and interaction service can not be awared and distinguished by trust models.2:The semantic relations between different trust categories and between different interact entities are not considered in these trust models, while it may lead to low efficiency and low accuracy.3:Most existed trust models are too complicated, and are not appropriate for resource constrained environments. These above problems are deeply studied and partly solved by new trust models proposed in this paper, based on this, we also list some research work on applying trust models under the pervasive environments. The applications of trust models for pervasive environments are concentrated on two aspects. The first research aspect is to take use of trust model to enhance trust levels between entities and increase the cooperation level between entities, related to this, two methods are studied in chapter5and chapter6to increase the cooperation level under different networks in pervasive environments. The second research aspect is to afford security guarantees for the applications in pervasive environments, related to this, a routing algorithm based on trust routing tree for WSN is proposed in chapter7. The main work and contributions are presented in the following aspects:(1) A Context-Aware and Self-Adaptive Trust Model(CASATM) for pervasive environments is proposed, our trust model provide security protection based on service content adaptively at various level, and be sensitive to contexts such as time, location and service content. A simple and efficient risk evaluation model was presented to initialize trust value to unknown entities which is a common phenomenon in pervasive environments. Simulation results testify to the contexts Awareness of our model, as well as resistance of periodically cheating behaviour from malicious entity.(2) A semantic distance based trust model is proposed in this paper. The semantic distance between entities and between trust categories is borrowed to calculate trustworthiness more precisely. In our model, the behavior trust and capability trust is distinguished, all entities in pervasive environments make independent decisions which can maximize their own profit with our trust model. The simulation experiment results proved the effectiveness of our model in increasing the interaction success ratio between entities as well as the interaction efficiency.(3) The application of distributed trust model in cooperation evolution under pervasive environments is researched. Game theory is borrowed and integrated in trust models to stimulate nodes’cooperation behavior. A distributed trust model based on a non-cooperative game that uses a genetic algorithm to let the nodes quickly learn the appropriate cooperation behavior is proposed in this paper, unlike traditional distributed model, a good behavior strategy with high fitness owns a higher transmitting efficiency to accelerate the convergence process. Further more, to prolong the network life, we expand the payoff table to encourage node’s selfish behavior in case of low energy level.(4) The application of trust model in cooperation evolution under multi-agents system of P2P networks is studied. An interest-trust based cooperation model is proposed, the interest-trustworthiness of agents is evaluated in different interest fields by utilizing Bayesian method. By updating the interest-trust based connection between agents, a robust multi-agent cooperative system is formed. Simulation results prove that the proposed interest-trust based cooperation model can not only increase the ratio of successful execution of tasks, but also improve network topology to cluster nodes around the trustworthy nodes in different interest fields. The advantage of our model include following features:a kind of interest-trust based clustering is formed, network assimilation is avoided and agent’s personality is embodied.(5) The application of trust model in the data packets routing under WSN networks is studied in chapter7. A novel routing algorithm base on trustworthy core tree (TCTR) is proposed aims to prolong network lifetime as well as increase network security in a hierarchical-cluster sensornet. Cluster heads with higher residual energy and trust level are elected from underlying sensor nodes. A minimum pathloss tree algorithm is borrowed to organize all cluster heads as a trustworthy core tree with sink node as tree root. Expand the Trustworthy core tree to cover all nodes so that each node report to sink node with a certain route. A trust model is integrated in TCTR to evaluate node’s trust level and detect evil nodes. Simulation results testify to the effectiveness of the algorithm in producing a longer network lifetime and a safer network.更多还原