Privacy Vulnerabilities in Synthetic Single-Cell RNA-Sequence Data
Golob, S.; McKeever, P.; Pentyala, S.; De Cock, M.; Peck, J.
Show abstract
MotivationSingle-cell RNA-Sequence (scRNA-seq) data is kept under strict access control because of its sensitive nature. Synthetic data generation (SDG) techniques can generate artificial scRNA-seq data, begging the question of whether leading SDG techniques like scDesign2 can sufficiently protect privacy of donors to justify relaxing access control. ResultsWe present an adversarial privacy attack against scDesign2, demonstrating that the leading technique for synthetic scRNA-seq data generation does not mask which donors were used to train the generator. The fewer the number of cell donors used in the dataset to train the SDG model, the easier an attack is. Using our experimental results, we argue for best practices to safely release synthetic data, mitigating privacy risks. Availability and Implementationhttps://github.com/steveng9/scRNA-seq_privacy_audits
Matching journals
The top 8 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- The Effect of Kinship in Re-identification Attacks Against Genomic Data Sharing Beacons 96%
- Accessibility of covariance information creates vulnerability in Federated Learning frameworks 96%
- Privacy-Preserving and Robust Watermarking on Sequential Genome Data using Belief Propagation and Local Differential Privacy 95%
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.