A Robust Privacy-Preserving Federated Learning Model Against Model Poisoning Attacks
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
Although federated learning offers a level of privacy by aggregating user data without direct access, it remains inherently vulnerable to various attacks, including poisoning attacks where malicious actors submit gradients that reduce model accuracy. In addressing model poisoning attacks, existing defense strategies primarily concentrate on detecting suspicious local gradients over plaintext. However, detecting non-independent and identically distributed encrypted gradients poses significant challenges for existing methods. Moreover, tackling computational complexity and communication overhead becomes crucial in privacy-preserving federated learning, particularly in the context of encrypted gradients. To address these concerns, we propose a robust privacy-preserving federated learning model resilient against model poisoning attacks without sacrificing accuracy. Our approach introduces an internal auditor that evaluates encrypted gradient similarity and distribution to differentiate between benign and malicious gradients, employing a Gaussian Mixture Model and Mahalanobis Distance for byzantine-tolerant aggregation. The proposed model utilizes Additive Homomorphic Encryption to ensure confidentiality while minimizing computational and communication overhead. Our model demonstrates superior performance in accuracy and privacy compared to existing strategies and encryption techniques, such as Fully Homomorphic Encryption and Two-Trapdoor Homomorphic Encryption. The proposed model effectively addresses the challenge of detecting maliciously encrypted non-independent and identically distributed gradients with low computational and communication overhead.
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 | 0.000 |
| 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 | 0.005 |
| Open science | 0.003 | 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