Research map / Adversarial robustness and security / Attacks on text and graphs
Simulated Adversarial Testing of Face Recognition Models
Nataniel Ruiz, Adam Kortylewski, Weichao Qiu, Cihang Xie, Sarah Adel Bargal, Alan Yuille, Stan Sclaroff
Check 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 |
|---|---|---|---|
| Use policy-gradient adversarial testing instead of uniform random sampling | CBAM-ResNet50 with ArcFace · Synthetic face verification, 15 shape and 15 texture parameters | accuracy: 99% -> 36% | Hurt |
| Use policy-gradient adversarial testing instead of uniform random sampling | CBAM-ResNet50 with ArcFace · Synthetic face verification, 15 shape and 15 texture parameters | Avg. Cosine Similarity: 0.518 -> 0.263 | Hurt |
| Use policy-gradient adversarial testing rather than random sampling or random optimization | Face recognition models · Synthetic face verification, 30 shape, 30 texture, and 2 pose parameters | accuracy: ArcFace CBAM-ResNet50: 99.9% -> 49%; CosFace CBAM-ResNet50: 99.9% -> 57%; ArcFace SE-CBAM-ResNet50: 99.8% -> 34%; ArcFac | Hurt |
| Filter negative-yaw images from training | CBAM-ResNet50 with ArcFace · Synthetic face verification, yaw [-1.0, -0.5] | Area Between Curves: 8.69 | Hurt |
| Filter positive-yaw images from training | CBAM-ResNet50 with ArcFace · Synthetic face verification, yaw [0.5, 1.0] | Area Between Curves: 8.46 | Hurt |
| Test with adversarial simulator parameters instead of random sampling | ArcFace CBAM-ResNet50 · Synthetic face verification, 30 shape, 30 texture, and 2 pose parameters | adversarial samples found: one adversarial sample over 1,000 iterations -> 51 adversarial samples over 12,587 iterations | Improved |
Closest papers by meaning
- Improving Transferability of Adversarial Patches on Face Recognition With Generative Models (CVPR 2021 · similarity 0.87)
- Evading Forensic Classifiers With Attribute-Conditioned Adversarial Faces (CVPR 2023 · similarity 0.87)
- Identifying Model Weakness with Adversarial Examiner (AAAI 2020 · similarity 0.87)
- Non-Adaptive Adversarial Face Generation (NeurIPS 2025 · similarity 0.85)
- HyperFace: Generating Synthetic Face Recognition Datasets by Exploring Face Embedding Hypersphere (ICLR 2025 · similarity 0.85)
- UIFace: Unleashing Inherent Model Capabilities to Enhance Intra-Class Diversity in Synthetic Face Recognition (ICLR 2025 · 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.
