Privacy-friendly Synthetic Data for the Development of Face Morphing Attack Detectors

TL;DR

This paper introduces the first synthetic dataset for face morphing attack detection, demonstrating its effectiveness in training models that generalize well to unknown attacks, while addressing legal challenges of real biometric data sharing.

Contribution

It presents a novel synthetic dataset (SMDD) for MAD development and shows its success in training high-performing detectors, overcoming legal and data sharing issues.

Findings

Synthetic data enables effective MAD training.

Models trained on SMDD perform well on unseen attack types.

SMDD dataset is publicly available for research.

Abstract

The main question this work aims at answering is: "can morphing attack detection (MAD) solutions be successfully developed based on synthetic data?". Towards that, this work introduces the first synthetic-based MAD development dataset, namely the Synthetic Morphing Attack Detection Development dataset (SMDD). This dataset is utilized successfully to train three MAD backbones where it proved to lead to high MAD performance, even on completely unknown attack types. Additionally, an essential aspect of this work is the detailed legal analyses of the challenges of using and sharing real biometric data, rendering our proposed SMDD dataset extremely essential. The SMDD dataset, consisting of 30,000 attack and 50,000 bona fide samples, is publicly available for research purposes.