Latent Dynamics · 2026-11-13

Context-Predictive Low-Rank Adaptation for Face Presentation Attack Detection

Aminu Yusuf, Shakeel Muhammad Ibrahim, Sulaimon Oyeniyi Adebayo, Muhammad Sohail Ibrahim, Muhammad Shahid Jabbar, Shujaat Khan

arXiv:2611.16534PDF

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

CP-JEPA adapts a pretrained CLIP (ViT-B/16) visual encoder for face presentation attack detection using low-rank adaptation (LoRA) plus context-predictive learning. A student predicts complete-image teacher features at masked spatial locations (EMA teacher targets) while retaining unmasked classification, masked classification, and prediction consistency. Inference uses only the LoRA-adapted CLIP visual encoder and binary classifier.

Why it matters

The authors address cross-dataset PAD transfer by explicitly supervising how local facial evidence is represented under partial observation, rather than only fine-tuning a final binary decision. CP-JEPA combines LoRA-based adaptation with masked latent prediction under a supervised PAD objective and reports improvements on standard OCIM and transfer scenarios, including unseen-attack evaluation on SiW-Mv2.

Method

  • LoRA-adapt CLIP ViT-B/16: trainable low-rank updates applied to query and value projections in all 12 transformer blocks, with pretrained CLIP weights frozen.
  • Context-predictive masked latent prediction: an EMA teacher provides normalized spatial targets for masked patch locations; a 2-layer transformer predictor outputs predicted latent features for masked positions.
  • Multi-objective supervised PAD training: unmasked CE, masked CE, and a KL-based prediction-consistency objective between unmasked and masked-view predictions; auxiliary components are training-only.

Limitation

The results show that context-predictive adaptation can improve cross-dataset PAD performance, although the gains vary with the training sources and evaluation setting.

Abstract (from arXiv)

Face presentation attack detection requires representations that distinguish physical spoofing cues across changes in acquisition devices, subjects, and attack media. Large visual encoders offer transferable representations, but conventional binary fine-tuning supervises their final decisions without explicitly constraining how local facial evidence is represented under partial observation. We present CP-JEPA, a context-predictive adaptation framework that combines low-rank tuning of a pretrained CLIP image encoder with a joint-embedding predictive objective. A student predicts complete-image teacher features at masked spatial locations, while unmasked classification, masked classification, and prediction consistency retain the connection to the live/spoof task. An exponential-moving-average teacher supplies evolving latent targets. The auxiliary components are used only during training; inference requires only the LoRA-adapted CLIP visual encoder and binary classifier. {We evaluate CP-JEPA under single-source and multi-source OCIM transfer, additional CelebA-Spoof training, transfer to CelebA-Spoof, two-source training, and unseen-attack evaluation on SiW-Mv2.} On the standard OCIM benchmark, CP-JEPA achieves 3.74\% mean HTER and 98.73\% mean AUC. {Single-source evaluation achieves 11.36\% mean HTER across twelve source--target combinations.} With CelebA-Spoof added to the three OCIM training sources, CP-JEPA reduces mean HTER from 3.45\% to 2.35\% and increases mean AUC from 98.68\% to 99.32\%. On SiW-Mv2, it achieves 3.36\% mean HTER across the held-out attack categories. Controlled loss ablations examine the contributions of latent prediction, masked classification, and prediction consistency. The results show that context-predictive adaptation can improve cross-dataset PAD performance, although the gains vary with the training sources and evaluation setting.

Related papers

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-11-12 · Indexed by Pub-AI: 2026-11-13 UTC

preprint

Context-Predictive Low-Rank Adaptation for Face Presentation Attack Detection

Aminu Yusuf, Shakeel Muhammad Ibrahim, Sulaimon Oyeniyi Adebayo, Muhammad Sohail Ibrahim, Muhammad Shahid Jabbar, Shujaat Khan

Preprint posted: 2026-11-12 · Original record ↗

arxiv
2611.16534
Explicit source versions (1)
  • v1 · 2026-11-12
  1. Preprint available

    Source date: 2026-11-12 · First detected: 2026-11-13 UTC · Historical baseline

    Source evidence (3)
Abstract · arxiv

Face presentation attack detection requires representations that distinguish physical spoofing cues across changes in acquisition devices, subjects, and attack media. Large visual encoders offer transferable representations, but conventional binary fine-tuning supervises their final decisions without explicitly constraining how local facial evidence is represented under partial observation. We present CP-JEPA, a context-predictive adaptation framework that combines low-rank tuning of a pretrained CLIP image encoder with a joint-embedding predictive objective. A student predicts complete-image teacher features at masked spatial locations, while unmasked classification, masked classification, and prediction consistency retain the connection to the live/spoof task. An exponential-moving-average teacher supplies evolving latent targets. The auxiliary components are used only during training; inference requires only the LoRA-adapted CLIP visual encoder and binary classifier. {We evaluate CP-JEPA under single-source and multi-source OCIM transfer, additional CelebA-Spoof training, transfer to CelebA-Spoof, two-source training, and unseen-attack evaluation on SiW-Mv2.} On the standard OCIM benchmark, CP-JEPA achieves 3.74\% mean HTER and 98.73\% mean AUC. {Single-source evaluation achieves 11.36\% mean HTER across twelve source--target combinations.} With CelebA-Spoof added to the three OCIM training sources, CP-JEPA reduces mean HTER from 3.45\% to 2.35\% and increases mean AUC from 98.68\% to 99.32\%. On SiW-Mv2, it achieves 3.36\% mean HTER across the held-out attack categories. Controlled loss ablations examine the contributions of latent prediction, masked classification, and prediction consistency. The results show that context-predictive adaptation can improve cross-dataset PAD performance, although the gains vary with the training sources and evaluation setting.