Privacy-preserving variation graphs

Pangenome graphs summarize the genomes of many individuals, so sharing them openly risks leaking private information about the people whose sequences built them. The NLnet-funded privvg project ("privacy-preserving variation graphs") addressed this by developing practical differential privacy models for variation graphs, work that was first prototyped in vg and has now landed in odgi as odgi priv. odgi priv applies the exponential mechanism to sample shared sub-haplotypes from the graph, and the strongest ε-differential privacy guarantees are obtained by publishing only the FASTA sequences of these sampled paths via odgi paths -f.

Acknowledgments

This project received funding through the NGI0 Discovery Fund, a fund established by the NLnet Foundation with financial support from the European Commission's Next Generation Internet programme under grant agreement No 825322.