Research map / Label-efficient and robust learning / Recognition and long-tail
Adversarially Learned Representations for Information Obfuscation and Inference
Martin Bertran, Natalia Martinez, Afroditi Papadaki, Qiang Qiu, Miguel Rodrigues, Galen Reeves, Guillermo Sapiro
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 |
|---|---|---|---|
| tighten tolerance k from 0.5 to 0.2 (smaller allowed leakage) (ablation) | adversarial filter · CelebA emotion, fixed estimator | emotion accuracy: 68.4% -> 51.9% | Improved |
| tighten tolerance k from 0.5 to 0.2 (ablation) | adversarial filter · CelebA gender detection | gender accuracy: 89.3% -> 83.9% | Hurt |
| filter versus blacking out lower face | adversarial filter k=0.2 · CelebA gender detection | gender accuracy: 75.5% -> 83.9% | Improved |
| filter versus blacking out full face | adversarial filter k=0.5 · CelebA gender detection | gender accuracy: 67.9% -> 89.3% | Improved |
| tighten tolerance k from infinity to 0.5 (ablation) | adversarial filter · FaceScrub consenting users | top-5 accuracy: 98.7% -> 97.6% | No clear effect |
| retrain final dense layer of subject network (ablation) | adversarial filter k=0.5 · FaceScrub subject verification | top-5 accuracy: 93.5% -> 96.8% | Improved |
| retrain final dense layer of subject network (ablation) | adversarial filter k=0.1 · FaceScrub subject verification | top-5 accuracy: 74.5% -> 89.6% | Improved |
| adversarially trained versus natural-image emotion estimator | adversarial filter k=0.2 · CelebA emotion | emotion accuracy: 51.9% -> 74.3% | Mixed |
| nonlinear versus linear filter | stochastic neural network filter · synthetic data tradeoff | I(U;X|Y) vs I(S;Y) tradeoff | Improved |
Closest papers by meaning
- Trade-offs and Guarantees of Adversarial Representation Learning for Information Obfuscation (NeurIPS 2020 · similarity 0.88)
- Mitigating Information Leakage in Image Representations: A Maximum Entropy Approach (CVPR 2019 · similarity 0.87)
- Learning Robust and Privacy-Preserving Representations via Information Theory (AAAI 2025 · similarity 0.86)
- Adversarial Machine Unlearning (ICLR 2025 · similarity 0.84)
- Perturb a Model, Not an Image: Towards Robust Privacy Protection via Anti-Personalized Diffusion Models (NeurIPS 2025 · similarity 0.84)
- Rethinking Adversarial Robustness in the Context of the Right to be Forgotten (ICML 2024 · similarity 0.83)
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.
