Research map / Adversarial robustness and security / Attacks on text and graphs
Adversarial Image Perturbation for Privacy Protection -- A Game Theory Perspective
Seong Joon Oh, Mario Fritz, Bernt Schiele
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 |
|---|---|---|---|
| add vaccination against blur (/B) to GAMAN (ablation) | GAMAN · PIPA val1 GoogleNet, recogniser B | recognition rate (%): 22.2 -> 5.8 | Improved |
| add noise vaccination (/N) to GAMAN (ablation) | GAMAN · PIPA val1 GoogleNet, recogniser N | recognition rate (%): 15.0 -> 4.6 | Improved |
| add noise vaccination (/N) to GAMAN (ablation) | GAMAN · PIPA val1 GoogleNet, recogniser B | recognition rate (%): 22.2 -> 23.6 | Hurt |
| use combined vaccination /TNBC instead of /B (ablation) | GAMAN · PIPA val1 GoogleNet, recogniser B | recognition rate (%): 5.8 -> 9.5 | Hurt |
| use gradient vector instead of gradient sign | FGV vs FGS · PIPA val1 AlexNet | recognition rate (%): 23.6 -> 13.3 | Improved |
| use gradient vector instead of gradient sign (multi-iteration) | GA vs BI · PIPA val1 AlexNet | recognition rate (%): 1.2 -> 0.2 | Improved |
| use multi-iteration instead of single-iteration gradient | GA vs FGV · PIPA val1 AlexNet | recognition rate (%): 13.3 -> 0.2 | Improved |
| continue past decision boundary with fixed steps (GAMAN) instead of early stopping (DF) | GAMAN vs DF · PIPA val1 GoogleNet, recogniser Proc | recognition rate (%): 51.8 -> 4.0 | Improved |
| add basic resize and quantisation processing against GA | GA · PIPA val1 GoogleNet | recognition rate (%): 0.0 -> 8.6 | Hurt |
| crop and resize applied against GAMAN | GAMAN · PIPA val1 ResNet152 | recognition rate (%): 0.0 -> 31.8 | Hurt |
| play randomised mixed strategy instead of deterministic /B | user strategy · PIPA val1 GoogleNet game | recognition rate (%): 8.6 -> 7.3 | Improved |
| user lacks knowledge of recogniser strategy N | /B user strategy · PIPA val1 GoogleNet game | recognition rate (%): 5.8 -> 8.6 | Hurt |
Closest papers by meaning
- Can Protective Perturbation Safeguard Personal Data from Being Exploited by Stable Diffusion? (CVPR 2024 · similarity 0.85)
- Perturb a Model, Not an Image: Towards Robust Privacy Protection via Anti-Personalized Diffusion Models (NeurIPS 2025 · similarity 0.85)
- PerceptAnon: Exploring the Human Perception of Image Anonymization Beyond Pseudonymization for GDPR (ICML 2024 · similarity 0.84)
- Prompt-Agnostic Adversarial Perturbation for Customized Diffusion Models (NeurIPS 2024 · similarity 0.84)
- PubDef: Defending Against Transfer Attacks From Public Models (ICLR 2024 · similarity 0.83)
- Revisiting Adversarial Patches for Designing Camera-Agnostic Attacks against Person Detection (NeurIPS 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.
