mlsanityData to 14 Oct 2026

Steering Vision-Language Pre-trained Models for Incremental Face Presentation Attack Detection

arXiv 2512.19022 ↗First posted 22 Dec 2025cs.CVcited by 0

On the map: Multimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness

Abstract (arXiv)

Face Presentation Attack Detection (PAD) demands incremental learning (IL) to combat evolving spoofing tactics and domains. Privacy regulations, however, forbid retaining past data, necessitating rehearsal-free IL (RF-IL). Vision-Language Pre-trained (VLP) models, with their prompt-tunable cross-modal representations, enable efficient adaptation to new spoofing styles and domains. Capitalizing on this strength, we propose SVLP-IL, a VLP-based RF-IL framework that balances stability and plasticity via textitMulti-Aspect Prompting (MAP) and textitSelective Elastic Weight Consolidation (SEWC). MAP isolates domain dependencies, enhances distribution-shift sensitivity, and mitigates forgetting by jointly exploiting universal and domain-specific cues. SEWC selectively preserves critical weights from previous tasks, retaining essential knowledge while allowing flexibility for new adaptations. Comprehensive experiments across multiple PAD benchmarks show that SVLP-IL significantly reduces catastrophic forgetting and enhances performance on unseen domains. SVLP-IL offers a privacy-compliant, practical solution for robust lifelong PAD deployment in RF-IL settings.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in MLLM jailbreaks and adversarial robustness, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 35 of its 54 references that are in the map. First posted on arXiv 22 Dec 2025.

Builds on

35 of 54 references.

References · 35

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  3. 2016Overcoming catastrophic forgetting in neural networkscited by 4,787
  4. 2021The Power of Scale for Parameter-Efficient Prompt Tuningcited by 3,795
  5. 2016Learning without Forgettingcited by 2,496
  6. 2015Learning Deep Features for Discriminative Localizationcited by 2,436
  7. 2022Visual Prompt Tuningcited by 1,681
  8. 2021Learning to Prompt for Continual Learningcited by 944
  9. 2023A Comprehensive Survey of Continual Learning: Theory, Method and Applicationcited by 872
  10. 2021DyTox: Transformers for Continual Learning with DYnamic TOken eXpansioncited by 291
  11. 2022S-Prompts Learning with Pre-trained Transformers: An Occam's Razor for Domain Incremental Learningcited by 246
  12. 2018Efficient parametrization of multi-domain deep neural networkscited by 168
23 more
  1. 2018Learning Deep Models for Face Anti-Spoofing: Binary or Auxiliary Supervisioncited by 167
  2. 2020Single-Side Domain Generalization for Face Anti-Spoofingcited by 139
  3. 2020CelebA-Spoof: Large-Scale Face Anti-Spoofing Dataset with Rich Annotationscited by 109
  4. 2022Domain Generalization via Shuffled Style Assembly for Face Anti-Spoofingcited by 73
  5. 2021Deep Learning for Face Anti-Spoofing: A Surveycited by 65
  6. 2025SD-LoRA: Scalable Decoupled Low-Rank Adaptation for Class Incremental Learningcited by 63
  7. 2017On Quadratic Penalties in Elastic Weight Consolidationcited by 55
  8. 2021Self-Domain Adaptation for Face Anti-Spoofingcited by 49
  9. 2021Generalizable Representation Learning for Mixture Domain Face Anti-Spoofingcited by 46
  10. 2020Revisiting Pixel-Wise Supervision for Face Anti-Spoofingcited by 45
  11. 2022Multi-domain Learning for Updating Face Anti-spoofing Modelscited by 39
  12. 2021Detection and Continual Learning of Novel Face Presentation Attackscited by 33
  13. 2022Generative Domain Adaptation for Face Anti-Spoofingcited by 33
  14. 2022Adaptive Mixture of Experts Learning for Generalizable Face Anti-Spoofingcited by 28
  15. 2021Feature Generation and Hypothesis Verification for Reliable Face Anti-Spoofingcited by 23
  16. 2022Digital and Physical Face Attacks: Reviewing and One Step Furthercited by 17
  17. 2025Componential Prompt-Knowledge Alignment for Domain Incremental Learningcited by 13
  18. 2023Rehearsal-Free Domain Continual Face Anti-Spoofing: Generalize More and Forget Lesscited by 12
  19. 2024CP-Prompt: Composition-Based Cross-modal Prompting for Domain-Incremental Continual Learningcited by 11
  20. 2022VLAD-VSA: Cross-Domain Face Presentation Attack Detection with Vocabulary Separation and Adaptationcited by 7
  21. 2019A database for face presentation attack using wax figure facescited by 6
  22. 2024Confidence Aware Learning for Reliable Face Anti-spoofingcited by 5
  23. 2024Generalized Face Liveness Detection via De-fake Face Generatorcited by 4

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2023FLIP: Cross-domain Face Anti-spoofing with Language Guidance
  2. 2025InstructFLIP: Exploring Unified Vision-Language Model for Face Anti-spoofing
  3. 2026CPG-PAD: Concept-Informed Prompts Guided Presentation Attack Detection
  4. 2025SLIP: Spoof-Aware One-Class Face Anti-Spoofing with Language Image Pretraining
  5. 2026Benchmarking Vision Foundation Models for Domain-Generalizable Face Anti-Spoofing
  6. 2025FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models
  7. 2025Exploring Task-Solving Paradigm for Generalized Cross-Domain Face Anti-Spoofing via Reinforcement Fine-Tuning
  8. 2026Foundation and Multimodal Large Language Models for Face Presentation and Morph Attack Detection