机会网络数据转发中的灰色关联与协同计算信任模型

发布时间：2018-04-09 09:35

本文选题：机会网络　切入点：信任模型　出处：《湖南科技大学》2017年硕士论文

【摘要】：机会网络是一种不需要源节点和目的节点之间存在完整路径,利用节点移动带来的相遇机会实现网络通信的自组织网络。而机会网络中如果存在自私节点或恶意节点,将对网络的性能和安全构成重大威胁。考虑上述恶劣环境,为了提高节点的传输成功率、减少传输延迟、提高网络的安全性,本文提出了一种基于灰色关联和协同计算的数据转发信任模型GCTM(Gray correlation and Collaborative computing Trust Model)。相关的工作主要有以下两个方面:第一,本文针对复杂环境下的机会网络特性,提出了基于灰色关联分析的直接信任值的计算。此模块通过分析节点的移动特性以及普通节点与自私、恶意节点间的差异,选取接触亲密度、投递可信度和位置亲密度作为模型的多维信任参数,计算节点的直接信任值,并采用灰色关联算法调节直接信任值的权重系数。此举将静态的权重系数改进为可根据信任参数间的内在联系自动调节各信任参数的动态权重系数,该方法提高了机会网络信任模型直接信任值计算的精确度。第二,本文针对机会网络中自私与恶意节点的攻击特性以及机会网络中节点的稀疏性,提出了一种基于协同计算的过滤算法来计算推荐信任值。该算法利用数据发送节点与邻居节点合作进行协同计算。采用接触亲密度、投递可信度作为计算推荐信任的参数。推荐信任模块运行时,评价节点向其周围邻居节点发送协同计算请求,邻居节点收到请求后将推荐信息发送给评价节点,考虑到网络中存在部分自私与恶意节点,所以需对推荐信息进行过滤。本文中采用KNN算法对推荐信息进行过滤处理。KNN算法可有效识别自私节点和恶意节点所发送的虚假推荐信任,并进行滤除,该方法减轻了自私与恶意节点对机会网络的干扰。采用协同计算技术,多节点联合计算推荐信任值,使计算结果更具可靠性。为了检验GCTM信任值计算模型的有效性,本文使用ONE仿真器进行仿真实验。实验中,通过在机会网络中设置不同的自私与恶意节点个数来验证模型的安全性。实验结果表明,本文提出的GCTM模型下的数据转发算法与经典的Epidemic、MaxProp、Direct Delivery算法以及MDT信任模型相比,其传输成功率、传输延迟、路由开销具有较好的改进。
[Abstract]:Opportunistic network is an ad hoc network which does not need a complete path between the source node and the destination node and uses the encounter opportunity brought by the node movement to realize the network communication.However, if there are selfish or malicious nodes in opportunistic networks, they will pose a serious threat to the performance and security of the networks.Considering the bad environment mentioned above, in order to improve the transmission success rate, reduce the transmission delay and improve the security of the network, this paper presents a data forwarding trust model GCTM(Gray correlation and Collaborative computing Trust model based on gray correlation and cooperative computing.The main work of this paper is as follows: first, aiming at the characteristics of opportunistic networks in complex environments, this paper presents the calculation of direct trust values based on grey correlation analysis.By analyzing the mobility characteristics of nodes and the differences between common nodes and selfish and malicious nodes, this module selects contact affinity, delivery credibility and location affinity as the multidimensional trust parameters of the model, and calculates the direct trust value of the nodes.Grey correlation algorithm is used to adjust the weight coefficient of direct trust value.In this way, the static weight coefficient is improved to automatically adjust the dynamic weight coefficients of each trust parameter according to the inherent relationship between trust parameters. This method improves the accuracy of direct trust value calculation in the trust model of opportunistic network.Secondly, aiming at the attacks of selfish and malicious nodes in opportunistic networks and the sparsity of nodes in opportunistic networks, a filtering algorithm based on cooperative computing is proposed to calculate the recommended trust values.The algorithm uses the data sending node to cooperate with the neighbor node for collaborative computation.Contact affinity and delivery reliability are used as parameters to calculate recommendation trust.When the recommendation trust module is running, the evaluation node sends the cooperative computing request to the neighboring node, and the neighbor node sends the recommendation information to the evaluation node after receiving the request, considering that there are some selfish and malicious nodes in the network.So it is necessary to filter the recommendation information.In this paper, the KNN algorithm is used to filter the recommendation information. KNN algorithm can effectively identify the false recommendation trust sent by the selfish node and the malicious node, and filter the false recommendation trust. This method reduces the interference of the selfish and malicious nodes to the opportunity network.The cooperative computing technique is used to calculate the recommended trust value jointly, which makes the calculation results more reliable.In order to verify the validity of the GCTM trust value calculation model, this paper uses the ONE simulator to carry on the simulation experiment.In the experiment, the security of the model is verified by setting different numbers of selfish and malicious nodes in the opportunistic network.Experimental results show that compared with the classical Epidemicus MaxPropDirect Delivery algorithm and the MDT trust model, the proposed data forwarding algorithm based on the GCTM model has a better improvement on the transmission success rate, transmission delay and routing overhead than that of the classical Epidemicus MaxPropDirect Delivery algorithm and the MDT trust model.
【学位授予单位】：湖南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.08

【参考文献】