SGAN-RA: Reconstruction Attack for Big Model in Asynchronous Federated Learning
Bibliographic record
Abstract
Federated learning (FL) is a distributed learning framework designed for large-scale applications. The core advantage of FL is that each participant is not required to share the local data with a central server. This inherent privacy-preserving capability is well suited to the increasingly popular large-scale generative AI models. However, some studies have shown that FL is susceptible to reconstruction attacks, in which an attacker leverages the acquired gradients or model parameters to reconstruct a victim's data. In this article, we propose a novel reconstruction attack scheme based on generative adversarial networks (GANs) in an asynchronous FL scenario that can reconstruct a victim's dataset without an auxiliary dataset. This adversarial scheme demonstrates a significant ability to reconstruct a dataset of victims accurately, thereby posing a substantial threat to user privacy, particularly in the context of large-scale models. Furthermore, we explore various defense mechanisms based on the characteristics of asynchronous FL and ultimately establish a viable defense scheme based on homomorphic encryption and an intermediate server. The proposed defense framework successfully and flawlessly defends against reconstruction attacks from the server side in an asynchronous setting without degradation of the model performance.
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 | 0.004 | 0.009 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.001 | 0.003 |
| Open science | 0.002 | 0.004 |
| Research integrity | 0.002 | 0.003 |
| 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".