Three reasons are provided in favour of $L^1$-norm as a measure ofprivacy-leakage: i) It is proved that this measure satisfies post-processingand linkage inequalities that make it consistent with an intuitive notion of aprivacy measure; ii) It is shown that the optimal utility-privacy trade-off canbe efficiently solved through a standard linear program when $L^1$-norm isemployed as the privacy measure; iii) It is also proved that it is sufficientto consider this measure of privacy in order to bound the privacy-leakagemeasured by mutual information, maximal leakage, or the improvement in aninference attack with a bounded cost function.
展开▼
