Microdata disclosure risk assessment using the family of conditional compound Poisson distributions

点击翻译按钮获取中文摘要

This project has progressed both on an empirical basis and on a theory. One empirical contribution is an anonymized data set of the Japanese Labor Force Survey, which will be used to test the risk and utility of anonymization methods. Another contribution is a survey on the advanced practices and researches of anonymization. This survey has been especially conducted to provide a sound view on the current Japanese anonymization practices and researches. On a theory, the limiting conditional compound Poisson distribution has been investigated to explain the behavior of a sparse contingency table, which often arises in assessing disclosure risk. In particular, a new asymptotics based on a Bell polynomial has been proposed.