Salta al contenuto principale
Passa alla visualizzazione normale.


Reputation Management in Distributed Systems


Several distributed applications, implemented over today’s Internet, are based on the assumption that participating agents collaborate in order to achieve their own goal. However, when these applications are modelled as unstructured distributed systems, the greater autonomy and decentralization encourage antisocial behaviours, which are likely to cause performance degradation for the whole system. This paper presents a fully distributed reputation management system that allows the evaluation of agent reputation in unstructured environments without any centralized coordination. The proposed approach is based on game theory and is capable of capturing the highly dynamic nature of the involved communities. As a representative example of an unstructured environment, Peer-to-peer (P2P) networks are considered. Those dynamic communities are affected by several antisocial behaviours, such as free riding. Since this phenomenon typically causes and exacerbates an unbalanced and unfair use of system resources, it has been considered as the case study in our work. The proposed solution exploits peer reputations in order to define an incentive system, whose main goal is the dissuasion from free riding.