P2P网络激励兼容的声誉评价及服务分配

由自私节点组成的P2P网络常常采用基于声誉的机制激励节点间的合作。当前已经提出了一类基于声誉的方案,但其中大部分方案都忽视了方案的分布式实现中惩罚的激励兼容性问题。在社会学或生态学领域,该问题常被称为二阶搭便车问题:拒绝低声誉节点的服务请求会降低惩罚者自身的声誉,因此对于自私节点而言,这种惩罚并不总是有利的选择。以P2P文件共享为例对该问题进行了分析,提出了一种连续型二阶声誉评价模型及两种服务分配规则,即概率分配规则和比例分配规则,实验表明只有概率分配规则是激励兼容的。

Incentive compatible reputation evaluation and service assignment in P2P networks

P2P networks are usually made of selfish nodes where the incentive mechanisms based on reputation are often used to induce cooperation. A class of schemes based on reputation has been proposed, but almost all of them ignore the incentive compatibility of punishment when the schemes are implemented in distributed model. It is usually called second-ordered free riding in the researches of sociology or ecology: refusing service request from nodes with low reputation will increase the punisher's own reputation, so refusing isn't always a favorable choice for selfish node. The paper discussed this problem in P2P file sharing networks and proposed a continuous second-ordered reputation evaluation model and two service assignment rules: probabilistic assignment rule and proportional assignment rule. The experiments show that only the first one is incentive compatible.