Research map / Adversarial robustness and security / Black-box and transfer attacks
ActiveThief: Model Extraction Using Active Learning and Unannotated Public Data
Soham Pal, Yash Gupta, Aditya Shukla, Aditya Kanade, Shirish Shevade, Vinod Ganapathy
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| active learning subset selection instead of random | ActiveThief · 6 datasets | agreement | Improved |
| DFAL+K-center instead of random | ActiveThief · GTSRB 30K budget | agreement: 83.84 -> 86.71 | Improved |
| uncertainty sampling instead of random | ActiveThief · image datasets | agreement | Mixed |
| combine DFAL with K-center (ablation) | ActiveThief DFAL · image datasets | agreement | Improved |
| natural public thief data instead of uniform noise | model extraction · CIFAR-10 | agreement: 10.62 -> 84.99 | Improved |
| more active learning iterations (10 -> 20) (ablation) | ActiveThief · MNIST 10K budget | agreement: 95.80 -> 96.74 | Improved |
| more active learning iterations (10 -> 20) (ablation) | ActiveThief · CIFAR-10 10K budget | agreement: 64.20 -> 64.23 | No clear effect |
| access to output probability scores (ablation) | ActiveThief · CIFAR-10 10K budget | agreement: 64.20 -> 77.29 | Improved |
| increase query budget | ActiveThief · 6 datasets | agreement | Improved |
| ActiveThief substitute instead of Papernot substitute | FGSM transfer · MNIST | transferability: 40.28 -> 59.55 | Improved |
| mismatched substitute architecture | ActiveThief · text datasets | agreement | No clear effect |
Closest papers by meaning
- Beyond Slow Signs in High-fidelity Model Extraction (NeurIPS 2024 · similarity 0.83)
- HoneypotNet: Backdoor Attacks Against Model Extraction (AAAI 2025 · similarity 0.83)
- CEGA: A Cost-Effective Approach for Graph-Based Model Extraction and Acquisition (ICML 2025 · similarity 0.82)
- Inconsistency-Based Data-Centric Active Open-Set Annotation (AAAI 2024 · similarity 0.82)
- SAME: Sample Reconstruction against Model Extraction Attacks (AAAI 2024 · similarity 0.82)
- From Counterfactuals to Trees: Competitive Analysis of Model Extraction Attacks (NeurIPS 2025 · similarity 0.82)
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.
