Trust management and matchmaking system for multi-attribute reverse auctions

在线拍卖欺诈占所有互联网欺诈行为的最大部分，由于拍卖容易受到机会主义行为的影响，在线拍卖欺诈行为正在上升。因此，拍卖被认为是信任关键系统。信任很难建立，因为交易发生在完全陌生的人之间。当今的拍卖面临着严重的欺诈行为，包括：竞价先令、狙击和掩护、拍卖人作弊、拍卖人-投标人作弊和缺乏隐私，尤其是落选的竞拍者。先令和拍卖商作弊是最难发现的欺诈类型。现有拍卖行报告的这些问题，使拍卖变得非常不公平和不可靠。为了提高拍卖的信任度，克服上述问题，我们希望开发一种可信任的在线拍卖行，用于多属性逆向拍卖，即在竞争卖家和多属性描述被拍卖产品的背景下。所有关于拍卖信任管理的现有工作都涉及前向拍卖，即与竞争买家的拍卖，以及简单且不能在许多现实世界中使用的仅限价拍卖。 我们值得信赖的多属性逆向拍卖行由以下几个部分组成： (1)解决狙击、屏蔽、拍卖商作弊和缺乏隐私问题的互联网拍卖新机制；(2)实时欺诈检测系统，在运行时检测两种作弊活动：先令和出价；(3)实时信任评估系统，动态计算卖家的信任分数；(4)评估拍卖协议公平性的框架；以及(5)匹配系统，搜索匹配买家请求的畅销产品。然后，具有可接受信任值的匹配卖家被邀请参加拍卖。我们将利用博弈论对协商协议的规则进行建模，并验证理想的理论性质、能够实现组件(2)和(4)活动的形式化方法，以及用于设计、实现和模拟我们的拍卖行的每个组件的多代理技术。

Online auction frauds, representing the largest part of all Internet frauds, are on the rise because auctions are vulnerable to opportunistic behaviors. Thus, auctions are considered trust-critical systems. Trust is difficult to establish because transactions occur among complete strangers. Nowadays auctions are facing serious frauds including: bid shilling, sniping and shielding, auctioneer cheating, auctioneer-bidder cheating and lack of privacy of submitted bids especially the losing ones. Shilling and auctioneer cheating are the hardest types of frauds to detect. These problems, reported in existing auction houses, make auctions really unfair and unreliable. To increase trust in auctions and overcome the problems listed above, we want to develop a trustworthy online auction house for multi-attribute reverse auctions, i.e. in the context of competing sellers and multiple attributes to describe the products being auctioned. All existing work on trust management for auctions concerns forward auctions, i.e. with competing buyers, and price-only auctions which are simple and cannot be used in many real-world situations. Our trustworthy multi-attribute reverse auction house is composed of the following components: (1) a new mechanism for Internet auctions that copes with sniping, shielding, auctioneer cheating and lack of privacy problems; (2) a real-time fraud detection system that detects at run time two cheating activities: shilling and outbidding; (3) a real-time trust evaluation system that dynamically computes the trust scores of sellers; (4) a framework that evaluates the fairness of our auction protocol, and (5) a matchmaking system that searches for the best sellers' products that match the buyer's request. The matched sellers who have acceptable trust values are then invited to the auction. We will utilize game theory to model the rules of our negotiation protocol and verify desirable theoretical properties, formal methods to be able to realize the activities of components (2) and (4), and multi-agent technology to design, implement and simulate each component of our auction house.