Privacy Preserving Distributed Computation of Hypertree Multiagent Organization
Kamala Srinivasan · The Atrium (University of Guelph) · 2013
Probabilistic reasoning, constraint reasoning, and decision theoretic reasoning are some of the essential tasks of multiagent systems. Many multiagent system frameworks exist to support these tasks. Some of these frameworks organize the agents into a structure called hypertree. The advantages of an hypertree organization include communication efficiency, inferential soundness and a high degree of agent privacy during normal inference operations. However, during hypertree organization construction, agent privacy may be compromised, and it is so for all existing hypertree based multiagent system frameworks. In this work, we propose distributed algorithms to recognize whether a hypertree organization exists for a given multiagent system, and if so, construct one. During the recognition and construction, our algorithms preserve agent privacy on private variables, shared variables, agent identities and bordering relations.