We present TrustDavis, an online reputation system that provides insurance against trade fraud by leveraging existing relationships between players, such as the ones present in social networks. Using TrustDavis and a simple strategy, an honest player can set an upper bound on the losses caused by any malicious collusion of players. In addition, TrustDavis incents participants to accurately rate each other, resists participants’ pseudonym changes, and is inherently distributed.
Some online auction sites have formalized the means by which individuals provide feedback on buyers and seller. Loosely speaking, we call such mechanisms online reputation systems: “A reputation system collects, distributes, and aggregates feedback about participants’ past behavior”. Examples are eBay’s Feedback Forum and the feedback ratings at overstock.com.
一部のオンラインオークションサイトでは、購入者と販売者に対してフィードバックする方法を公式に定めている。大まかに、そのようなメカニズムをオンラインレピュテーションシステム（参加者の過去の行動についてのフィードバックを収集・分配・集約するようなレピュテーションシステム）と呼ぶこととする。例としては、eBay の Feedback Forum や overstock.com の feedback ratings などが挙げられる。
Such systems usually assign a rating to a particular identity. Ideally, individuals with good ratings are reliable trade partners, whereas individuals with poor ratings should be avoided. Unfortunately, the reputation systems now available on the internet can be manipulated by malicious individuals or groups for selfish purposes. For example, a group can collude to artificially improve an individual’s ratings with the intent of tricking unsuspecting victims into trading with someone who will never deliver the goods. This is the well known “hit and run” problem, to which all unsecured bilateral exchange is susceptible as there is always the temptation to receive a good or service without reciprocation.
This problem is aggravated online as many trading partners are veiled by relative anonymity and rarely trade. Mechanisms that have been proposed to mitigate such problems have achieved limited success. Ideally, we want a reputation system that resists malicious manipulation by groups of colluding parties, or at least that provides a strategy that honest participants can use to limit their exposure to such manipulation. TrustDavis addresses this concern. Its properties are:
The outline of the paper is as follows. Section 2 briefly reviews of the current literature, focusing on motivating the three properties not yet discussed in detail. Section 3 describes the basic framework of the system, the use of references. Sections 3.1 and 3.2 obtain upper and lower bounds on the price of references. Section 4 describes a strategy that helps honest players avoid exploitation by malicious ones. In section 5, we provide suggestions for further research and in section 6 we summarize our results.