Research map / Federated learning and privacy / Privacy attacks and secure learning
SensitiveNets: Learning Agnostic Representations with Application to Face Images
Aythami Morales, Julián Fiérrez, Rubén Vera-Rodríguez, Rubén Tolosana
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 |
|---|---|---|---|
| apply SensitiveNets projection to embeddings | ResNet-50 face embeddings · LFW verification | accuracy: 98.4% -> 95.8% | Hurt |
| apply SensitiveNets projection | ResNet-50 embeddings · DiveFace gender (NN) | accuracy: 97.7% -> 58.8% | Improved |
| apply SensitiveNets projection | ResNet-50 embeddings · DiveFace ethnicity (NN) | accuracy: 98.8% -> 55.1% | Improved |
| apply SensitiveNets projection | ResNet-50 embeddings · DiveFace gender (SVM) | accuracy: 96.2% -> 56.3% | Improved |
| apply SensitiveNets projection | ResNet-50 embeddings · DiveFace gender (RF) | accuracy: 95.1% -> 54.6% | Improved |
| apply SensitiveNets projection | ResNet-50 embeddings · DiveFace ethnicity (SVM) | accuracy: 98.2% -> 54.1% | Improved |
| apply SensitiveNets projection | ResNet-50 embeddings · DiveFace ethnicity (RF) | accuracy: 97.3% -> 53.5% | Improved |
| obfuscate gender at image level with differential privacy | ResNet-50 embeddings · CelebA gender (SVM) | accuracy: 98.4% -> 98.3% | No clear effect |
| apply SensitiveNets rather than image-level obfuscation | ResNet-50 embeddings · CelebA gender (SVM) | accuracy: 98.4% -> 67.3% | Improved |
| train SVM on SensitiveNets representation | SVM on embeddings · CelebA Attractiveness | equal opportunity gap: 30.1 -> 6.8 | Improved |
| train SVM on SensitiveNets representation | SVM on embeddings · CelebA Smiling | equal opportunity gap: 14.5 -> 6.0 | Improved |
| train SVM on SensitiveNets representation | SVM on embeddings · CelebA Attractiveness | accuracy: 74.3 -> 77.7 | Improved |
Closest papers by meaning
- Attribute-Preserving Face Dataset Anonymization via Latent Code Optimization (CVPR 2023 · similarity 0.90)
- Enhancing Facial Privacy Protection via Weakening Diffusion Purification (CVPR 2025 · similarity 0.87)
- Privacy-Preserving Face Recognition Using Trainable Feature Subtraction (CVPR 2024 · similarity 0.86)
- Balancing Privacy and Performance: A Many-in-One Approach for Image Anonymization (AAAI 2025 · similarity 0.86)
- Disguise without Disruption: Utility-Preserving Face De-identification (AAAI 2024 · similarity 0.86)
- Transferable Facial Privacy Protection against Blind Face Restoration via Domain-Consistent Adversarial Obfuscation (ICML 2024 · similarity 0.85)
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.
