Differentially Private Data Release for Data Analytics - A Model Review

To leverage on the potential of data analytics, enabling private data release is needed. The challenge in achieving private data release has been balancing between privacy and analytical utility. Among the models that seek to solve the challenge, ε-differential privacy promises to achieve the balance by regulating the epsilon (ε) value. The choice of the appropriate epsilon value that achieves the balance has been a challenge, making the ε-differential privacy not practically applicable by many. A practical and heuristic method to estimate this privacy parameter needs formulation. The variable to estimate appropriate privacy parameter that is not provided in heuristic manner is the reidentification probability. Previous research has based that probability on released data sets and linkage data sets, with less focus on data analysts. This paper proposes a causal relationship model for estimating the reidentification probability, which adds the analyst's aspect to the model.