Generalized Analysis of Convergence of Absolute Trust in Peer-to-Peer Networks
To mitigate the attacks by malicious peers and to motivate the peers to share the resources in peer-to-peer networks, several reputation systems have been proposed in the past. In most of them, the peers evaluate other peers based on their past interactions and then aggregate this information in the whole network. However such an aggregation process requires approximations in order to converge at some global consensus. It may not be the true reflection of past behavior of the peers. Moreover such type of aggregation gives only the relative ranking of peers without any absolute evaluation of their past. This is more significant when all the peers responding to a query, are malicious. In such a situation, we can only know that who is better among them without knowing their rank in the whole network. In this paper, we are proposing a new algorithm which accounts for the past behavior of the peers and will estimate the absolute value of the trust of peers. Consequently, we can suitably identify them as a good peers or malicious peers. Our algorithm converges at some global consensus much faster by choosing suitable parameters. Because of its absolute nature it will equally load all the peers in network. It will also reduce the inauthentic download in the network which was not possible in existing algorithms.