Attacks on genetic privacy via uploads to genealogical databases.

Area of Science:

• Genetics
• Bioinformatics
• Computer Science

Background:

• Direct-to-consumer (DTC) genetic services are widely used, with millions of customers uploading their genetic data.
• Genealogy platforms often allow users to upload genetic data to find relatives based on shared genomic regions (identical by state - IBS).
• Existing security measures in these databases may not adequately protect against sophisticated data extraction.

Purpose of the Study:

• To investigate and demonstrate methods by which an adversary can infer genotypes from DTC genetic databases.
• To quantify the amount of genetic data an adversary can recover using uploaded datasets.
• To identify vulnerabilities in current IBS detection methods used by genealogy services.

Main Methods:

• Simulating an adversary uploading multiple genetic datasets to a DTC genealogy database.
• Utilizing identical by state (IBS) region detection algorithms with unphased genotypes.
• Performing genome-wide genetic imputation based on recovered allelic information.

Main Results:

• An adversary uploading approximately 900 genomes could recover at least one allele at SNP sites for up to 82% of a median European genome.
• Approximately 100 falsified uploads can yield sufficient genetic information for genome-wide imputation in databases using unphased genotypes.
• A proof-of-concept attack was successfully demonstrated on the GEDmatch database.

Conclusions:

• Current DTC genetic databases are vulnerable to genotype inference attacks by adversaries.
• The methods described pose a significant risk to genetic data privacy for users of genealogy services.
• Countermeasures are necessary to prevent such exploits and protect user genetic information.