Abstract: In the existing reputation models, the accumulation of reputation of certain node requires so long a period even if that node provides good services, which influences the positivity of those newly added nodes. Besides, most models composite the global trust value by multiple iteration, and the huge amount of iteration will bring massive computational expense. On account of the above issue, this dissertation proposes a new reputation computing model by introducing the typical push-pull model in streaming media scheduling strategy. The push model can accelerate the trust value accumulation speed of those newly added nodes which provide good services, while the pull model reduces the network flows online and avoids the negative impacts of the iteration computation. Through analysis and simulation, this model can ensure the accuracy of reputation computation, meanwhile, the communication and computation expenses can get improved as well.

Key words: reputation, peer, push-pull protocol, iteration

摘要： 在现有的信誉模型中，即使节点积极提供良好的服务，节点信誉的累积也需要一个很长的周期，影响了新节点加入网络的积极性。此外，大部分模型在合成全局信誉值时采用多次迭代的方式，大量的迭代运算将导致巨大的计算开销。针对上述问题，通过引入流媒体调度策略中典型的推拉模式，提出一个新的信誉计算模型。在推模式下，对于那些新加入且积极提供优质服务的节点，可以加快其信誉累积速度，在拉模式下，减少了网络消息流量，避免了迭代计算的负面影响。分析及仿真表明，该模型在保证信誉计算准确性的同时，能较大程度改善通信及计算开销。

关键词: 信誉, 节点, 推拉协议, 迭代