An anonymization technique for databases is proposed that employs Principal Component Analysis. The technique aims at releasing the least possible amount of information, while preserving the utility of the data released in response to queries. The general scheme is described, and alternative metrics are proposed to assess utility, based respectively on matrix norms; correlation coefficients; divergence measures, and quality indices of database images. This approach allows to properly measure the utility of output data and incorporate that measure in the anonymization method.
View on arXiv