Modeling SNP-Trait Associations and Realizing Privacy-Utility Tradeoff in Genomic Data Publishing

Abstract

Genomic data privacy arises as one of the most important concerns facing the wide commoditization of DNA-genotyping. In this paper, we study the problem of privacy preserved kin-genomic data publishing. The major challenge in protecting kin-genomic data privacy is to protect against powerful attackers with abundant background knowledge. We propose a probabilistic model based on factor graph with the knowledge of publicly available GWAS statistics to reveal the dependency relationship between genotypes and phenotypes. Furthermore, a genomic data sanitization method is proposed to protect against optimal inference attacks launched by powerful attackers.