Biologically-Informed Hybrid Membership Inference Attacks on Generative Genomic Models

The increased availability of genetic data has transformed genomics research, but raised many privacy concerns regarding its handling due to its sensitive nature. This work explores the use of language models (LMs) for the generation of synthetic genetic mutation profiles, leveraging differential privacy (DP) for the protection of sensitive genetic data. We empirically evaluate the privacy guarantees of our DP modes by introducing a novel Biologically-Informed Hybrid Membership Inference Attack (biHMIA), which combines traditional black box MIA with contextual genomics metrics for enhanced attack power. Our experiments show that both small and large transformer GPT-like models are viable synthetic variant generators for small-scale genomics, and that our hybrid attack leads, on average, to higher adversarial success compared to traditional metric-based MIAs.

• Publication year

  2025

• Venue

  arXiv.org

• Publication date

  2025-11-10

• Fields of study

  Biology, Computer Science

• Identifiers
  DOI 10.48550/arXiv.2511.07503arXiv 2511.07503
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Assessing Privacy Vulnerabilities in Genetic Data Sets: Scoping Review

  2024cited by this paper
• Opportunities for basic, clinical, and bioethics research at the intersection of machine learning and genomics

  2023cited by this paper
• Synthetic Data Revolutionizes Rare Disease Research: How Large Language Models and Generative AI are Overcoming Data Scarcity and Privacy Challenges

  2023cited by this paper
• Synthetic is all you need: removing the auxiliary data assumption for membership inference attacks against synthetic data

  2023cited by this paper
• Membership Inference Attacks on Machine Learning: A Survey

  2021cited by this paper
• Extracting Training Data from Large Language Models

  2020cited by this paper
• A comprehensive survey and analysis of generative models in machine learning

  2020cited by this paper
• A Survey of Privacy Attacks in Machine Learning

  2020cited by this paper
• Re‐identifiability of genomic data and the GDPR

  2019cited by this paper
• Language Models are Unsupervised Multitask Learners

  2019cited by this paper
• A brief history of bioinformatics

  2018influential reference
• ML-Leaks: Model and Data Independent Membership Inference Attacks and Defenses on Machine Learning Models

  2018cited by this paper
• Privacy Challenges of Genomic Big Data.

  2017cited by this paper
• Exposed! A Survey of Attacks on Private Data

  2017cited by this paper
• Membership Inference Attacks Against Machine Learning Models

  2016cited by this paper
• Deep Learning with Differential Privacy

  2016cited by this paper
• A global reference for human genetic variation

  2015influential reference
• Quantifying Genomic Privacy via Inference Attack with High-Order SNV Correlations

  2015cited by this paper
• Privacy in the Genomic Era

  2014influential reference
• Bayesian method to predict individual SNP genotypes from gene expression data

  2012cited by this paper
• Genetic exceptionalism

  2010cited by this paper
• Sequence analysis Advance Access publication June 7, 2011 The variant call format and VCFtools

  2010cited by this paper
• Resolving Individuals Contributing Trace Amounts of DNA to Highly Complex Mixtures Using High-Density SNP Genotyping Microarrays

  2008influential reference
• Advances on natural language processing

  2007cited by this paper
• The Human Genome Project: Lessons from Large-Scale Biology

  2003cited by this paper
• Information

  2001cited by this paper
• Pangenome-Informed Language Models for Privacy-Preserving Synthetic Genome Sequence Generation

  year unknowncited by this paper

• No citing papers are available for this paper.