Research map / Adversarial robustness and security / Backdoor and poisoning
Detection and Continual Learning of Novel Face Presentation Attacks
Mohammad Rostami, Leonidas Spinoulas, Mohamed Hussein, Joe Mathai, Wael Abd-Almageed
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 GMM-based out-of-training-distribution (OTDS) detection with a third class output (ablation) | full replay with real labels · PADISI-Face (single PA/task) | APCER (%): 75.2 -> 13.0 | Improved |
| add GMM-based OTDS detection (task 1) (ablation) | full replay with real labels · PADISI-Face (single PA/task) | ACER (%): 37.3 -> 33.7 | Improved |
| use true labels for novel samples instead of model-predicted labels (reduce label pollution) (ablation) | NACL · PADISI-Face (single PA/task) | ACER (%): 17.8 -> 8.3 | Improved |
| use true labels for novel samples instead of predicted labels (ablation) | NACL · PADISI-Face (single PA/task) | BPCER (%): 17.4 -> 2.8 | Improved |
| use true labels for novel samples instead of predicted labels (ablation) | NACL · PADISI-Face (single PA/task) | APCER (%): 18.2 -> 13.8 | Improved |
| present attacks from difficult to easy rather than easy to difficult (ablation) | NACL · PADISI-Face (single PA/task) | ACER (%): 19.2 -> 24.9 | Hurt |
| present attacks from difficult to easy rather than easy to difficult (ablation) | NACL · PADISI-Face (single PA/task) | APCER (%): 24.9 -> 35.3 | Hurt |
| replay all stored samples (full replay) instead of 100 selected samples near cluster means | NACL · PADISI-Face (single PA/task) | polluted labels (%) at task 8: 9.8 -> 9.3 | No clear effect |
| select buffer samples closest to GMM means instead of random selection (ablation) | NACL · PADISI-Face (single PA/task) | ACER | Improved |
| learn two novel attack types per task instead of one | NACL · PADISI-Face / HQ-WMCA | BPCER | Improved |
Closest papers by meaning
- Rehearsal-Free and Efficient Continual Learning for Cross-Domain Face Anti-Spoofing (TPAMI 2025 · similarity 0.87)
- Multi-Domain Incremental Learning for Face Presentation Attack Detection (AAAI 2024 · similarity 0.86)
- One-Class Face Anti-spoofing via Spoof Cue Map-Guided Feature Learning (CVPR 2024 · similarity 0.85)
- SLIP: Spoof-Aware One-Class Face Anti-Spoofing with Language Image Pretraining (AAAI 2025 · similarity 0.85)
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models (AAAI 2026 · similarity 0.84)
- Test-Time Domain Generalization for Face Anti-Spoofing (CVPR 2024 · similarity 0.84)
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.
