Decentralized information dissemination in multidimensional semantic social overlays

Abstract

Information dissemination in a decentralized network is a challenging issue as no single node can have knowledge of the whole network for reasons like security, access control, memory. Most importantly - a centralized service provider is missing which can cater to user requests. In this paper we propose and investigate an algorithm for information dissemination in decentralized networks which only uses local information - that is information associated with a node and its neighbors for disseminating a piece of information with a goal of maximum coverage and minimum spamming. The approach emulates multiagent system using particle swarms, and explores multiple relations that exist among nodes while selecting potential forwarders in case members having same interests are not well connected. The approach is generic enough to be used in various kinds of scenarios: for example in job market or network of emails, decentralized recommendation and contextual advertisement systems. We measure various metrics such as recall, precision and message duplication to analyze the quality of our protocol. The experimental results on real and synthetic dataset show the effectiveness of our approach. © 2012 Springer-Verlag.