Research map / Adversarial robustness and security / Black-box and transfer attacks
Unravelling Robustness of Deep Learning Based Face Recognition Against Adversarial Attacks
Gaurav Goswami, Nalini Ratha, Akshay Agarwal, Richa Singh, Mayank Vatsa
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 grid-based occlusion | VGG-Face · PaSC | GAR (%) at 1% FAR: 54.3 -> 3.2 | Hurt |
| Add most-significant-bit noise | VGG-Face · PaSC | GAR (%) at 1% FAR: 54.3 -> 1.3 | Hurt |
| Add eye-region occlusion | VGG-Face · MEDS-II | GAR (%) at 1% FAR: 78.4 -> 10.9 | Hurt |
| Add grid-based occlusion | LightCNN · MEDS-II | GAR (%) at 1% FAR: 89.3 -> 80.1 | Hurt |
| Add xMSB noise | L-CSSE · MEDS-II | GAR (%) at 1% FAR: 89.1 -> 83.4 | Hurt |
| Use hidden-layer response distances with an SVM detector | Face recognition networks · PaSC, proposed image distortions | detection accuracy (%) | Improved |
| Use hidden-layer response distances with an SVM detector | Face recognition networks · MEDS-II, Grids | detection accuracy (%): 68.4 | Mixed |
| Use only the top 6 hidden-layer features for detection (ablation) | VGG-Face detector · MEDS-II, average across distortions | detection accuracy (%): 81.7% | Mixed |
| Use only the single most discriminative hidden-layer feature (ablation) | VGG-Face detector · MEDS-II, average across distortions | detection accuracy (%): 79.3% | Mixed |
| Apply selective dropout and median filtering to detected distorted images | LightCNN · PaSC | GAR (%) at 1% FAR: 25.9 -> 36.2 | Improved |
| Apply selective dropout and median filtering to detected distorted images | LightCNN · MEDS-II | GAR (%) at 1% FAR: 41.6 -> 61.3 | Improved |
| Apply selective dropout and median filtering to detected distorted images | VGG-Face · PaSC | GAR (%) at 1% FAR: 14.6 -> 24.8 | Improved |
Closest papers by meaning
- FaceSec: A Fine-Grained Robustness Evaluation Framework for Face Recognition Systems (CVPR 2021 · similarity 0.89)
- Efficient Decision-Based Black-Box Adversarial Attacks on Face Recognition (CVPR 2019 · similarity 0.89)
- Exploring Architectural Ingredients of Adversarially Robust Deep Neural Networks (NeurIPS 2021 · similarity 0.88)
- Harnessing Edge Information for Improved Robustness in Vision Transformers (AAAI 2024 · similarity 0.86)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.86)
- Robust NAS under adversarial training: benchmark, theory, and beyond (ICLR 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.
