Private Collaborative Data Cleaning via Non-Equi PSI
Bibliographic record
Abstract
We introduce and investigate the privacy-preserving version of collaborative data cleaning. With collaborative data cleaning, two parties want to reconcile their data sets to filter out badly classified, misclassified data items. In the privacy-preserving (private) version of data cleaning, the additional security goal is that parties should only learn their misclassified data items, but nothing else about the other party’s data set. The problem of private data cleaning is essentially a variation of private set intersection (PSI), and one could employ recent circuit-PSI techniques to compute misclassifications with privacy. However, we design, analyze, and implement three new protocols tailored to the specifics of private data cleaning that outperform a circuit-PSI-based approach. With the first protocol, we exploit the idea that a small additional leakage (the differentially private size of the intersection of data items) allows for a reduction in complexity over circuit-PSI. The other two protocols convert the problem of finding a mismatch in data classifications into finding a match, and then follow the standard technique of using oblivious pseudorandom functions (OPRF) for computing PSI. Depending on the number of data classes, this leads to a concrete runtime improvement over circuit-PSI.
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 | 0.010 | 0.028 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.002 |
| Bibliometrics | 0.001 | 0.003 |
| Science and technology studies | 0.002 | 0.004 |
| Scholarly communication | 0.004 | 0.012 |
| Open science | 0.006 | 0.014 |
| Research integrity | 0.003 | 0.005 |
| Insufficient payload (model declined to judge) | 0.005 | 0.002 |
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".