REPUTATION MANAGEMENT ALGORITHMS IN DISTRIBUTED APPLICATIONS

Nowadays, several distributed systems and applications rely on interactions between unknown agents that cooperate in order to exchange resources and services. The distributed nature of these systems, and the consequent lack of a single centralized point of control, let agents to adopt selfish and malicious behaviors in order to maximize their own utility. To address such issue, many applications rely on Reputation Management Systems (RMSs) to estimate the future behavior of unknown agents before establishing actual interactions. The relevance of these systems is even greater if the malicious or selfish behavior exhibited by a few agents may reduce the utility perceived by cooperative agents, leading to a damage to the whole community. RMSs allow to estimate the expected outcome of a given interaction, thus providing relevant information that can be exploited to take decisions about the convenience of interacting with a certain agent. Agents and their behavior are constantly evolving and becoming even more complex, so it is increasingly difficult to successfully develop the RMS, able to resist the threats presented. A possible solution to this problem is the use of agent-based simulation software designed to support researchers in evaluating distributed reputation management systems since the design phase. This dissertation presents the design and the development of a distributed simulation platform based on HPC technologies called DRESS. This solution allows researchers to assess the performance of a generic reputation management system and provides a comprehensive assessment of its ability to withstand security attacks. In the scientific literature, a tool that allows the comparison of distinct RMS and different design choices through a set of defined metrics, also supporting large-scale simulations, is still missing. The effectiveness of the proposed approach is demonstrated by the application scenario of user energy sharing systems within smart-grids and by considering user preferences differently from other work. The platform has proved to be useful for the development of an energy sharing system among users, which with the aim of maximizing the amount of energy transferred has exploited the reputation of users once learned their preferences.