Preserving confidentiality of high-dimensional tabular data: Statistical and computational issues