Heterogeneous Differential-Private Federated Learning: Trading Privacy for Utility Truthfully
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Differential-private federated learning (DP-FL) has emerged to prevent privacy leakage when disclosing encoded sensitive information in model parameters. However, the existing DP-FL frameworks usually preserve privacy homogeneously across clients, while ignoring the different privacy attitudes and expectations. Meanwhile, DP-FL is hard to guarantee that uncontrollable clients (i.e., stragglers) have truthfully added the expected DP noise. To tackle these challenges, we propose a heterogeneous differential-private federated learning framework, named HDP-FL, which captures the variation of privacy attitudes with truthful incentives. First, we investigate the impact of the HDP noise on the theoretical convergence of FL, showing a tradeoff between privacy loss and learning performance. Then, based on the privacy-utility tradeoff, we design a contract-based incentive mechanism, which encourages clients to truthfully reveal private attitudes and contribute to learning as desired. In particular, clients are classified into different privacy preference types and the optimal privacy-price contracts in the discrete-privacy-type model and continuous-privacy-type model are derived. Our extensive experiments with real datasets demonstrate that HDP-FL can maintain satisfactory learning performance while considering different privacy attitudes, which also validate the truthfulness, individual rationality, and effectiveness of our incentives.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.004 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 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 it