Optimal Utility-Privacy Trade-off with the Total Variation Distance as the Privacy Measure
share
Three reasons are provided in favour of L^1-norm as a measure of privacy-leakage: i) It is proved that this measure satisfies post-processing and linkage inequalities that make it consistent with an intuitive notion of a privacy measure; ii) It is shown that the optimal utility-privacy trade-off can be efficiently solved through a standard linear program when L^1-norm is employed as the privacy measure; iii) It is also proved that it is sufficient to consider this measure of privacy in order to bound the privacy-leakage measured by mutual information, maximal leakage, or the improvement in an inference attack with a bounded cost function.
READ FULL TEXT
