Improved Detection of Face Presentation Attacks Using Image Decomposition

TL;DR

This paper introduces a face presentation attack detection method leveraging image decomposition to extract albedo and normal maps, enhancing domain gap analysis and achieving state-of-the-art results across multiple datasets.

Contribution

The paper proposes a novel PAD algorithm using image decomposition and contrastive loss to improve spoof detection accuracy and domain gap analysis.

Findings

Achieves state-of-the-art results on multiple face spoofing datasets.

Demonstrates the effectiveness of albedo and normal maps in PAD.

Shows the importance of lighting and contrast effects in spoof detection.

Abstract

Presentation attack detection (PAD) is a critical component in secure face authentication. We present a PAD algorithm to distinguish face spoofs generated by a photograph of a subject from live images. Our method uses an image decomposition network to extract albedo and normal. The domain gap between the real and spoof face images leads to easily identifiable differences, especially between the recovered albedo maps. We enhance this domain gap by retraining existing methods using supervised contrastive loss. We present empirical and theoretical analysis that demonstrates that contrast and lighting effects can play a significant role in PAD; these show up, particularly in the recovered albedo. Finally, we demonstrate that by combining all of these methods we achieve state-of-the-art results on both intra-dataset testing for CelebA-Spoof, OULU, CASIA-SURF datasets and inter-dataset setting on SiW, CASIA-MFSD, Replay-Attack and MSU-MFSD datasets.