Reputation System for Decentralized Population Census

We describe a framework and techniques for running decentralized census processes that enable observers to independently verify governmental data. Census is a process impacting important issues such as the representation of a community in the Congress and the amount of funding that it gets from a central government. Correct census is essential for detecting vote stuffing. Reliable census data can enable user certification for addressing fake identities and Sybil attacks. Census has been historically run by governments, but citizens and NGOs need to be able to verify it. Classical census is expensive and beyond the reach of these players, hence the need for affordable citizen-driven census technology. Various citizens have different opinions as to what information should be gathered and what makes a person eligible to be accounted for in statistics. Using as inputs the official preferences of a given government should enable the verification of the data of that government. The reported work formalizes this problem and introduces a framework for reasoning about census data. An addressed challenge is how to quantify the uncertainty and the trust in the data provided by users. We report on techniques to reason and to extract census-related conclusions based on the available data. Probabilistic models with various approximations are experimented for evaluating the census results in this context.