The use of concept hierarchies in privacy preserving data acquisition for data mining
Bibliographic record
Abstract
This thesis presents a concept hierarchy-based approach to privacy preserving data collection for data mining called the p-level model. The p-level model allows data providers to divulge information at any chosen privacy level (p-level), on any attribute. Data collected at a high p-level signifies divulgence at a higher conceptual level and thus ensures more privacy. Data providers have greater control of their privacy preferences, and have provided significantly (25-75%) more personal data values, at various p-levels, than when providing the same information using the regular, fixed-level ( f-level) method of data collection. However, the data mining process, which involves the integration of various data values, can constitute a privacy breach if combinations of attributes at the various p-levels result in the inference of knowledge that exists at lower p-levels. Providing anonymity guarantees prior to release can further protect the collected data set from privacy breaches due to linking the released data set with external data sets. This thesis describes the p-level reduction phenomenon and proposes methods to identify and control the occurrence of this privacy breach. One objective of this thesis is to explore the feasibility of applying data collected with the p-level approach to data mining problems. We apply data collected using the p-level approach to a data classification problem, and discover that the mining accuracy of the p-level approach classifier is comparable to that of the f-level (no privacy) approach, thus we conclude that the p-level approach is beneficial for the purpose of privacy preserving data collection.
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 | 0.024 | 0.050 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.003 |
| Bibliometrics | 0.004 | 0.006 |
| Science and technology studies | 0.003 | 0.008 |
| Scholarly communication | 0.007 | 0.015 |
| Open science | 0.003 | 0.007 |
| Research integrity | 0.001 | 0.006 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
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".