Privacy Homomorphisms for Subcontracting Statistical Computation

When publishing data containing official statistics, a need to preserve statistical confidentiality arises. Statistical disclosure of individuals’ data must be prevented. There is a wide choice of techniques to achieve this anonymization: data perturbation, data suppression, etc. In this paper, we tackle the problem of using anonymized data to compute exact statistics; the goal is for a classified level (statistical institute) to be able to retrieve statistics computed by an unclassified level (external contractor) on disclosure-protected macrodata. Our approach is based on privacy homomorphisms, especially on a recent one.