Abstract:
|
Primary suppression of small, or otherwise 'at risk' cells, is only the first step in producing disclosure safe aggregate tables. The next, and often more complex, task is the selection of cells for complementary suppression to ensure that the sensitive values are truly protected from disclosure. This task becomes more complex when there are numerous multidimensional tables with overlapping dimensions, as there is potential that suppression patterns used in one table may reveal information that compromises the suppressions in another table. We present a methodology that centralizes the selection and management of all suppressed cells when producing multiple tables from the same data source. This methodology also presents the opportunity to use log-linear modeling to estimate the values of suppressed cells using the values of suppressed cells and marginal totals as predictors. This allows the measurement of the disclosure risk for those values protected via suppression.
|