Assessing the Effectiveness of Singling-Out Attacks on Synthetic Data: Is the Current GDPR Guidance Adequate?
Bibliographic record
Abstract
Privacy remains one of the central challenges to data sharing under contemporary data protection regulations. To mitigate privacy risks, real datasets are either sanitized using anonymization techniques such as generalization or suppression, or replaced with artificially generated synthetic data. The European General Data Protection Regulation (GDPR) defines a singling-out attack as the ability to isolate specific records and identify an individual based on the content of a released dataset. Data controllers must ensure that published data does not allow singling out. However, this concept was originally designed for anonymized datasets, where a singled-out record corresponds to a real individual. In contrast, synthetic datasets contain artificially generated records, which may or may not represent or coincide with real individuals. This distinction raises questions about the applicability and effectiveness of singling-out attacks on synthetic data. To assess the effectiveness of singling-out attacks, we evaluated the attack on synthetic data generated by three well-known generative models (CTGAN, PATEGAN, and TabDDPM) across four real-world datasets. Our study reveals inherent limitations of conventional singling-out attacks when applied to synthetic data. As the GDPR’s privacy risk framework was originally developed with anonymized data in mind rather than synthetic data, there is a critical need for further research to establish robust evaluation frameworks and metrics specifically designed to assess the privacy risks associated with synthetic data.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.030 | 0.121 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.004 |
| Scholarly communication | 0.003 | 0.006 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.003 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".