Privacy-preserving Chi-squared testing for genome SNP databases

In recent years, the importance of privacy protection in genome-wide association studies (GWAS) has been increasing. GWAS focuses on identifying single-nucleotide polymorphisms (SNPs) associated with certain diseases such as cancer and diabetes, and Chi-squared testing can be used for this. However, recent studies reported that publishing the p-value or the corresponding chi-squared value of analyzed SNPs can cause privacy leakage. Several studies have been proposed for the anonymization of the chi-squared value with differential privacy, which is a de facto privacy metric in the cryptographic community. However, they can be applied to only small contingency tables; otherwise, they lose a lot of useful information. We propose novel anonymization methods: Rand-Chi and RandChiDist, and these methods are experimentally evaluated using real data sets.