Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SRL-MAD: Structured Residual Latents for One-Class Morphing Attack Detection

Diogo J. Paulo, Hugo Proença, João C. Neves

Abstract

Face morphing attacks represent a significant threat to biometric systems as they allow multiple identities to be combined into a single face. While supervised morphing attack detection (MAD) methods have shown promising performance, their reliance on attack-labeled data limits generalization to unseen morphing attacks. This has motivated increasing interest in one-class MAD, where models are trained exclusively on bona fide samples and are expected to detect unseen attacks as deviations from the normal facial structure. In this context, we introduce SRL-MAD, a one-class single-image MAD that uses structured residual Fourier representations for open-set morphing attack detection. Starting from a residual frequency map that suppresses image-specific spectral trends, we preserve the two-dimensional organization of the Fourier domain through a ring-based representation and replace azimuthal averaging with a learnable ring-wise spectral projection. To further encode domain knowledge about where morphing artifacts arise, we impose a frequency-informed inductive bias by organizing spectral evidence into low, mid, and high-frequency bands and learning cross-band interactions. These structured spectral features are mapped into a latent space designed for direct scoring, avoiding the reliance on reconstruction errors. Extensive evaluation on FERET-Morph, FRLL-Morph, and MorDIFF demonstrates that SRL-MAD consistently outperforms recent one-class and supervised MAD models. Overall, our results show that learning frequency-aware projections provides a more discriminative alternative to azimuthal spectral summarization for one-class morphing attack detection.

SRL-MAD: Structured Residual Latents for One-Class Morphing Attack Detection

Abstract

Face morphing attacks represent a significant threat to biometric systems as they allow multiple identities to be combined into a single face. While supervised morphing attack detection (MAD) methods have shown promising performance, their reliance on attack-labeled data limits generalization to unseen morphing attacks. This has motivated increasing interest in one-class MAD, where models are trained exclusively on bona fide samples and are expected to detect unseen attacks as deviations from the normal facial structure. In this context, we introduce SRL-MAD, a one-class single-image MAD that uses structured residual Fourier representations for open-set morphing attack detection. Starting from a residual frequency map that suppresses image-specific spectral trends, we preserve the two-dimensional organization of the Fourier domain through a ring-based representation and replace azimuthal averaging with a learnable ring-wise spectral projection. To further encode domain knowledge about where morphing artifacts arise, we impose a frequency-informed inductive bias by organizing spectral evidence into low, mid, and high-frequency bands and learning cross-band interactions. These structured spectral features are mapped into a latent space designed for direct scoring, avoiding the reliance on reconstruction errors. Extensive evaluation on FERET-Morph, FRLL-Morph, and MorDIFF demonstrates that SRL-MAD consistently outperforms recent one-class and supervised MAD models. Overall, our results show that learning frequency-aware projections provides a more discriminative alternative to azimuthal spectral summarization for one-class morphing attack detection.
Paper Structure (12 sections, 5 equations, 2 figures, 2 tables)

This paper contains 12 sections, 5 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Residual Fourier Map
  5. Ring-Based 2D Feature Representation
  6. Learnable Ring Projection
  7. Frequency-Band Refinement
  8. Autoencoder with Template Reconstruction
  9. Training Objective
  10. Experimental Setup
  11. Results
  12. Conclusion

Figures (2)

  • Figure 1: Illustration of SRL-MAD. From a single face image, residual Fourier features are extracted and divided into low, mid, and high-frequency bands. These are projected into a compact 1D latent space where morphing attacks appear well-separated from bona fides.
  • Figure 2: Overview of the proposed method (SRL-MAD). We extract the Residual Fourier Map from a single face image by subtracting a power-law baseline to highlight morphing artifacts. The map is divided into concentric rings to form a structured 2D feature representation, which is then processed through a learnable ring-wise spectral projection mechanism. These features are refined into low, mid, and high-frequency bands and projected by an autoencoder into a compact 1D latent space. The final 1D latent coordinate serves as a discriminative score.