文摘
In this paper, we consider how the security of a disclosure control mechanism based on randomised, but uncontrolled, rounding can be assessed by Bayesian methods. We develop a methodology, based on Markov chain Monte Carlo, for estimating the conditional (posterior) probability distribution for the original cell counts given the released rounded values. An effective rounding-based disclosure control will result in high posterior uncertainty about the true value. Conversely, a posterior distribution concentrated on a single value provides evidence of ineffective disclosure control.