Perennial secure multi-party computation of universal Turing machine

Consider a user who needs to perform computation over an initially unbounded stream of input. The user would like to compute functions of the input that change dynamically due to external events. The focus of this work is outsourcing such computation to a set of agents. The outsourcing must meet several constraints. Any large enough subset of agents must correctly emulate the user's computation on the unbounded stream of input. Any small subset of agents must obtain as little information as possible on the user's data, including the computed functions and any initial input. This privacy assurance must be maintained in an information-theoretic sense. Finally, the set of agents is dynamic with agents joining and leaving the set and different sets of agents being merged, cloned or split.