The increasing use of wireless infrastructure and portable technologies, such as laptops, cellular phones, and Personal Digital Assistants (PDAs), has enabled nodes to connect to wireless networks whenever access is available or to form spontaneous ad hoc networks to exchange files or context information.
This new communication paradigm changes the role of the user from consumer to content provider and it affects the way self-organizing systems, like peer to peer networks, are used. Furthermore, device mobility and user mobility create new interesting services designed specifically for nomadic users.
Peer to peer networks need to have self-organization properties because of the lack of a centralized authority. This implies that nodes should self-manage and cooperate to sustain the availability of the resources in the system. In this context, researchers have suggested the use of reputation management schemes to foster cooperation and to mitigate the attacks of misbehaving nodes.
In this thesis, we analyze the principles of autonomic communication system to define new security requirements that targets malicious and selfish nodes. We evaluate the basic components of reputation management systems and study the cost vs. the benefit of using reputation in peer to peer applications.
Borrowing techniques from game theory, we discuss the importance for a node to build and use its reputation value. We propose a framework, based on the generalized form of the Iterated Prisoner's Dilemma, to model the interactions of rational and selfish nodes in distributed systems. We study how a node takes into account the change of its reputation when deciding its behaviour in a transaction and discuss the Nash equilibrium in the system.
Then, we estimate the performance of reputation management schemes for specific applications. We simulate the capability of these schemes in reducing inauthentic and corrupted file transfers in end-user collaborative content-distribution systems.
Finally we define a new mechanism to improve the application of reputation management schemes in mobile scenarios. The token-based approach enables the correlation of reputation values earned in different contexts and virtual communities. This results in reducing the problem of reputation bootstrap in new communities and in accelerating the identification of malicious nodes.
Keywords: Autonomic systems, Reputation management, Game theory, Mobility