Dynamic Self-management of Autonomic Systems: The Reputation, Quality and Credibility (RQC) Scheme

In this paper, we present a feedback-based system for managing trust and detecting malicious behavior in autonomically behaving networks. Like other distributed trust management systems, nodes rate the interactions they have with other nodes and this information is stored in a distributed fashion. Two crucial insights motivate our work. We recognize as separate entities the trust placed in a node, reputation, and the trust placed in the recommendations made by a node, credibility. We also introduce the concept of quality of a trust rating. Together, these two factors enhance the ability of each node to decide how much confidence it can place in a rating provided to it by a third party. We implement our scheme on a structured P2P network, Pastry, though our results can be extended to generic autonomic communication systems. Experimental results considering different models for malicious behavior indicate the contexts in which the RQC scheme performs better than existing schemes.