Figure 1:

An overview of the proposed scheme. In (A), researchers coordinate to decide on the set of SNPs (I) that will be shared with the server and a common seed which is used to generate the permutation vec-tor Q. (B) shows the generation of the metadata Mⁱ from the original dataset Dⁱ belonging to the researcher Rⁱ. Initially, the researcher adds n^′ synthetic samples and later shuffles the SNPs (columns) based on the permutation in vector Q (e.g., the values for the first SNP are moved to the last column as shown by the cyan colored arrow). (B) also shows that some SNP values/states change for some users due to addition of noise to achieve the proposed LDP variant. The values in red denote the flipped SNP values. In (C), each researcher Rⁱ sends their prepared metadata Mⁱ to the server, which computes the pairwise kinship coefficients among all users and sends back to the researchers the related IDs.