Research map / Adversarial robustness and security / Attacks on text and graphs
Increasing the Cost of Model Extraction with Calibrated Proof of Work
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Apply privacy-cost-calibrated proof-of-work before releasing predictions | Public ML model queries · MNIST, SVHN, and CIFAR10 | query execution time: up to 2x for legitimate users; up to 100x for attackers | Mixed |
| Wrap predictions in proof-of-work while releasing the original, unmodified outputs | Victim model · MNIST, SVHN, and CIFAR10 | victim model accuracy | No clear effect |
| Use out-of-distribution or synthetic queries instead of legitimate in-distribution queries | Model extraction attacks · MNIST, SVHN, and CIFAR10 | proof-of-work execution time | Hurt |
| Minimize query entropy with the EntropyRev strategy | Model extraction attacks · MNIST, SVHN, and CIFAR10 | extracted-model accuracy | Hurt |
| Select queries with minimal victim privacy cost using WorstCase | Model extraction attacks · In-distribution queries | privacy cost | No clear effect |
| Combine out-of-distribution queries with occasional in-distribution queries | InOut attack · MNIST, SVHN, and CIFAR10 | privacy cost | Mixed |
| Use MixMatch semi-supervised extraction with unlabeled in-distribution examples | Model extraction attacks · MNIST, SVHN, and CIFAR10 | task accuracy | Helped |
Closest papers by meaning
- Exploring Query Efficient Data Generation Towards Data-Free Model Stealing in Hard Label Setting (AAAI 2025 · similarity 0.87)
- From Counterfactuals to Trees: Competitive Analysis of Model Extraction Attacks (NeurIPS 2025 · similarity 0.87)
- SAME: Sample Reconstruction against Model Extraction Attacks (AAAI 2024 · similarity 0.87)
- Stealix: Model Stealing via Prompt Evolution (ICML 2025 · similarity 0.87)
- HoneypotNet: Backdoor Attacks Against Model Extraction (AAAI 2025 · similarity 0.87)
- Reconstruction Attacks on Machine Unlearning: Simple Models are Vulnerable (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.