Reproduce “Biologically-Informed Hybrid Membership Inference Attacks on Generative Genomic Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Asia Belfiore, Jonathan Passerat-Palmbach, Dmitrii Usynin
The authors introduce a genomics-aware membership-inference attack to test whether a mutation profile trained a genomic generator. It combines model-response features with genomic features of a synthetic profile generated from one of the target’s mutations. Across chromosome 22 experiments, the hybrid generally performs better, especially without differential privacy, but results fluctuate and are limited to this study.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.