Design and Evaluation of Internet Auction Protocols robust against cheatings

本项目设计并研究了组合拍卖协议，该协议对Internet上的各种欺诈行为具有（几乎）鲁棒性。我们重点研究了Ad Hoc网络上的数据传输，例如P2P网络和分布式传感器网络。我们首先表明，现有的拍卖协议的影响，假名操纵。更准确地说，节点的所有者可以假装一个节点实际上是一组多个节点，或者所有这些节点都必须从不同的所有者那里购买。我们提出并分析了一个假名证明机制，其中所有者可能有多个元素。我们在WI-2007上展示了这些结果。另一方面，我们开发了基于Agent社区的对等信息检索（ACP 2 P）方法。该方法使用Agent社区来管理和查找用户感兴趣的信息。代理作为其用户的代表，通过与其他代理通信来搜索用户想要的信息。在WI-2007上发表了这些结果。此外，在回收市场、垄断中介市场等进行了多主体模拟和主体实验。虽然这些结果取决于特定的市场情况，但我们扩展了主体的决策模型，并成功地再现了多主体模拟中的主体行为。最后，作为我们研究结果的一个综合，我们研究了Internet上新机制的自动机制设计技术。传统的机制设计是针对一类问题进行手工设计的，而在自动机制设计中，可以使用约束优化技术来自动设计机制。我们的研究结果有望为实际的人类和大规模多智能体系统开发和研究更快，更好的机制。

This project designs and examines combinatorial auction protocols which are(almost)robust against a variety of cheatings on the Internet.We focused on a data transfer on ad hoc networks, such as P2P networks and distributed sensor networks. We first show that existing auction protocols are influenced by false-name manipulations. More precisely, an owner of nodes can pretend that a node is in fact a set of multiple nodes, or all of which must be purchased from different owners. We proposed and analyzed a false-name-proof mechanism, where an owner may have multiple elements. We presented those results at WI- 2007On the other hand, we developed the Agent-Community-based Peer-to-Peer Information Retrieval(ACP2P)method. This method uses agent communities to manage and look up information of interest to users. An agent works as a delegate of its user and searches for information that the user wants by communicating with other agents. We presented those results at WI-2007In addition, we conducted multi-agent simulations and subject experiments in recycle markets, monopolistic intermediary markets, and so on. Though those results depend on a specific market situation, we extended subject's decision-making models and successfully reproduce subject behavior in the multi-agent simulations. Also, we presented lots of papers in false-name-proof auction protocol and cooperative games in journals and conferences.Finally,as an integration of our results,we study automated mechanism design,as a mechanism design technique,for new mechanisms on the Internet Mechanism has traditionally been designed manually for classes of problems.In automated mechanism design,a mechanism can be automatically designed using constrained optimization technique. Our results are expected to develop and examine more rapidly and better mechanisms for actual human and massively multiagent systems.