EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Privacy-hardened and hallucination-resistant synthetic data generation with logic-solvers

Mark A Burgess, Brendan Hosking, Roc Reguant iD, Anubhav Kaphle, Mitchell J O'Brien, Letitia M F Sng iD, Yatish Jain, Denis C Bauer iD

Bioinformatics · 01.12.2025 · 41 (12)

DOI: 10.1093/bioinformatics/btaf600 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Bioinformatics, online erschienen am 04.11.2025. Verfasst von Mark A Burgess und 7 weiteren Autor:innen. Beteiligte Einrichtung: Australian e-Health Research Centre, Commonwealth Scientific and Industrial Research Organisation, Canberra, 2601, Australia. Hauptthemen laut MEDLINE: Genomics, Privacy. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 33 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: CSIRO.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Motivation Machine-generated or synthetic data is a valuable resource for training artificial intelligence algorithms, evaluating rare workflows, and sharing data under stricter data legislations. However, current statistical and deep learning methods struggle with large data volumes, are prone to hallucinating scenarios incompatible with reality, and seldom quantify privacy meaningfully. Results Here, we introduce Genomator, a logic solving approach (SAT solving), which efficiently produces private and realistic representations of the original data. We demonstrate the method on genomic data, which arguably is the most complex and private information. We benchmark Genomator against state-of-the-art methodologies (Markov generation, Wasserstein Generative Adversarial Network and Conditional Restricted Boltzmann Machines), demonstrating a 40%–530% accuracy improvement and 57%–172% higher privacy. Genomator is also 3–100 times more efficient, making it the only tested method that scales to whole genomes. We show the universal trade-off between privacy and accuracy, and use Genomator’s tuning capability to cater to all applications along the spectrum, from provable private representations of sensitive cohorts, to datasets with indistinguishable pharmacogenomic profiles. Demonstrating the production-scale generation of tuneable synthetic genomes hold great potential for balancing underrepresented populations in medical research and advancing global data exchange. Availability and implementation Genomator is available at https://github.com/csiro/genomator.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

GenomicsPrivacyAlgorithmsHumansArtificial Intelligence

Autor:innen und Institutionen

  • Australian e-Health Research Centre, Commonwealth Scientific and Industrial Research Organisation, Canberra, 2601, Australia.
  • Australian e-Health Research Centre, Commonwealth Scientific and Industrial Research Organisation, Sydney, 2145, Australia.
  • Australian e-Health Research Centre, Commonwealth Scientific and Industrial Research Organisation, Melbourne, 3052, Australia.
  • Applied BioSciences, Faculty of Science and Engineering, Macquarie University, Macquarie Park, 2109, Australia.
  • Australian e-Health Research Centre, Commonwealth Scientific and Industrial Research Organisation, Adelaide, 5000, Australia.
  • Department of Biomedical Informatics and Digital Health, School of Medical Sciences, University of Sydney, Sydney, 2050, Australia.

ORCID-Profile: Roc Reguant, Letitia M F Sng, Denis C Bauer

Förderung

  • CSIRO

Zitieren

APA

Burgess, M. A., Hosking, B., Reguant, R., Kaphle, A., O'Brien, M. J., Sng, L. M. F., Jain, Y., & Bauer, D. C. (2025). Privacy-hardened and hallucination-resistant synthetic data generation with logic-solvers. Bioinformatics, 41(12). https://doi.org/10.1093/bioinformatics/btaf600

Vancouver

Burgess MA, Hosking B, Reguant R, Kaphle A, O'Brien MJ, Sng LMF, et al. Privacy-hardened and hallucination-resistant synthetic data generation with logic-solvers. Bioinformatics. 2025;41(12). doi:10.1093/bioinformatics/btaf600

BibTeX
@article{Burgess2025,
  author = {Burgess, Mark A and Hosking, Brendan and Reguant, Roc and Kaphle, Anubhav and O'Brien, Mitchell J and Sng, Letitia M F and Jain, Yatish and Bauer, Denis C},
  title = {Privacy-hardened and hallucination-resistant synthetic data generation with logic-solvers},
  journal = {Bioinformatics},
  year = {2025},
  volume = {41},
  number = {12},
  doi = {10.1093/bioinformatics/btaf600}
}
Kostenlos starten →