presenters:

• Sam Joseph
• Zoran Despotovic

The challenges faced by a community of agents is very similar to that of a community of peers. One in particular is providing appropriate assurance mechanisms that reduce or eliminate opportunistic behaviour. Traditional forms of such mechanisms, such as contractual agreements monitored by enforcement institutions, are not always viable in an online world, particularly in global decentralized communities such as P2P networks. Under such circumstances, reputation systems that act as informal social mechanisms for encouraging trustworthy behaviour appear to be the only remaining alternative. Their key assumption is that previous actions are predictive of future behaviour. In this tutorial we deal with the problem of promoting trust in P2P systems by managing peers' reputations. We start with giving a formal problem definition. We then analyze a number of recently developed approaches and provide a classification of the possible solutions.

A strategy for aggregating trust-related feedback will:

Using these three dimensions to classify the existing approaches, the classes identified are: social networks, probabilistic estimation techniques and game-theoretic reputation systems. Social networks and probabilistic techniques target so called probabilistic behavior. Probabilistic techniques normally imply smaller implementation overheads than social networks and enable a more intuitive transition of reputation to trust, while social networks better detect misbehavior in a variety of settings. Game-theoretic models are suitable for rational behavior. However, decentralized game-theoretic reputation models are rather rare due to complexities stemming from the necessary decentralization of the reputation aggregation. Throughout, we draw parallels between peer to peer and agent trust mechanisms.