Mutually Private Verifiable Machine Learning As-a-service: A Distributed Approach
Bibliographic record
Abstract
Reliability is a crucial component to machine-learning-as-a-service platforms, as more and more critical applications depend on them. Thus, mechanisms employed to assure the integrity of computations performed on such platforms are pivotal to their robust functioning. Moreover, privacy protection, and performance guarantee at scale, are other major challenges surrounding these platforms that are by no means straightforward to overcome at the same time. In this paper, we have proposed a novel distributed approach, which uses specialized composable proof systems at its core, to respond to these challenges. At a high level, we adopt a divide-and-conquer approach to build efficient proof systems for machine-learning-based services in order to ensure the correctness of results. More precisely, the mathematical formulation of the machine learning task is divided into multiple parts, each of which is handled by a different specialized proof system; these proof systems are then combined with the commit-and-prove methodology to guarantee correctness as a whole. With privacy safeguards built into the design, our approach also assures that neither user data nor model parameters, which constitute the intellectual property of service providers are exposed in the process. We have showcased the usability of our approach within a machine learning service provider that offers classification services through a linear support vector machine (SVM) model. Our complexity analysis indicates that our system could be used in practical settings.
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 | 0.017 | 0.032 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.003 | 0.005 |
| Scholarly communication | 0.008 | 0.014 |
| Open science | 0.005 | 0.011 |
| Research integrity | 0.003 | 0.006 |
| Insufficient payload (model declined to judge) | 0.006 | 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".