Face morphing attacks are widely recognized as one of the most challenging threats to face recognition systems used in electronic identity documents. These attacks exploit a critical vulnerability in passport enrollment procedures adopted by many countries, where the facial image is often acquired without a supervised live capture process. In this paper, we propose a novel face morphing technique based on Arc2Face, an identity-conditioned face foundation model capable of synthesizing photorealistic facial images from compact identity representations. We demonstrate the effectiveness of the proposed approach by comparing the morphing attack potential metric on two large-scale sequestered face morphing attack detection datasets against several state-of-the-art morphing methods, as well as on two novel morphed face datasets derived from FEI and ONOT. Experimental results show that the proposed deep learning-based approach achieves a morphing attack potential comparable to that of landmark-based techniques, which have traditionally been regarded as the most challenging. These findings confirm the ability of the proposed method to effectively preserve and manage identity information during the morph generation process.
Figures & tables
Fig. 1 : Samples of morphed images (center column) generated with our proposed method. On the first row, Accomplice (left column) and criminal (right column) are sourced from the ONOT [ 10 ] dataset; on the second row, accomplice and criminal are sourced from the FEI [ 46 ] dataset.
Fig. 2 : Overview of the proposed method. The input images IA and IB are first encoded using the identity encoder Eid(⋅) , producing the identity embeddings eA and eB . In parallel, we extract the pose conditioning image for Arc2Face with EMOCAv2 [ 6 , 19 ] , obtaining IP . Then, the two compact identity representations eA and eB are mapped into the CLIP latent space and interpolated to obtain the latent representation pM , which is then decoded alongside with the conditioning image IP to generate the morphed image IM′ . Finally, the resulting image is post-processed using BEN2 to remove the background, yielding the final morphed image IM .
Algorithm
Type
FEI Morph v2
MONOT
EINMorph-HQ v2
EINMorph-MQ v2
C01 [ 41 ]
L
×
×
C02 [ 13 ]
L
×
×
C03 [ 43 ]
L
×
×
C05 [ 16 ]
L
×
×
C05-PA97 [ 16 ] [ 50 ] 1 1 1 Morphed images generated using the C05 approach [ 16 ] were subsequently retouched to remove morphing artifacts by applying a face restoration technique.
L
×
×
C05-PA98 [ 16 ] [ 49 ] 1
L
×
×
TABLE I: List of the morphing algorithms used to generate the morphed images for the datasets considered in this work. The column Type indicates whether the corresponding method is based on facial landmarks (L) or on deep learning techniques (D).
1
2
3
C01
97.9%
91.7%
74.0%
C02
99.7%
98.9%
95.9%
C03
97.3%
89.9%
70.2%
C05
98.4%
93.5%
78.0%
C08
98.6%
94.4%
82.5%
C15
97.7%
91.5%
70.9%
TABLE II : MAP computed on the FEI Morph v2 dataset for the proposed approach and seven competitors. Best results for each column are highlighted in bold. In this case, a single probe image is available, so the MAP of each algorithm is represented by a single row.
1
2
3
C01
94.2%
86.7%
73.5%
C02
94.1%
86.7%
74.0%
C03
90.9%
80.8%
63.4%
C05
91.8%
83.0%
67.4%
C15
89.3%
78.3%
55.8%
C16
91.8%
82.3%
61.6%
TABLE III : MAP computed on the MONOT dataset for the proposed approach and the six competitors reported in [ 3 ] , considering a single ONOT image for each subject as probe(the MAP of each algorithm is represented by a single row). Best results for each column are highlighted in bold.
1
2
3
1
100.0%
100.0%
99.9%
2
100.0%
100.0%
99.8%
3
100.0%
100.0%
99.8%
4
100.0%
100.0%
99.6%
5
100.0%
100.0%
99.6%
6
100.0%
99.9%
99.1%
TABLE IV : MAP computed on the MONOT dataset for the proposed approach, considering ten ”in the wild” images for each subject as probe.
Fig. 3 : Visualization of the MAPs computed on the MONOT dataset for the proposed approach and the competitors, considering ten ”in the wild” images as probe.
Fig. 4 : Visualization of the MAPs computed on the EINMorph-HQ v2 dataset for the proposed approach and the competitors.
Fig. 5 : Visualization of the MAPs computed on the EINMorph-MQ v2 dataset for the proposed approach and the competitors.
Fig. 6 : A visual example of the morphed images obtained using the different interpolation approaches, given the same image pair as input.
Dataset
Interp. location
finterp
MAPAvg
FEI
Identity
lerp
0.9778
slerp
0.9679
CLIP
lerp
0.9747
slerp
0.9835
MONOT
Identity
lerp
0.8539
slerp
0.8486
TABLE V : MAPAvg reported on FEI and MONOT datasets for each choice of interpolation location ( i.e. , identity-level or CLIP latent-level) and interpolation function finterp ( i.e. , linear interpolation or spherical linear interpolation).
Advancing face morphing attack techniques is crucial to anticipate evolving threats and develop robust defensive mechanisms for identity verification systems. This work introduces DCMorph, a dual-stream diffusion-based morphing framework that simultaneously operates at both identity conditioning and latent space levels. Unlike image-level methods suffering from blending artifacts or GAN-based approaches with limited reconstruction fidelity, DCMorph leverages identity-conditioned latent diffusion models through two mechanisms: (1) decoupled cross-attention interpolation that injects identity-specific features from both source faces into the denoising process, enabling explicit dual-identity conditioning absent in existing diffusion-based methods, and (2) DDIM inversion with spherical interpolation between inverted latent representations from both source faces, providing geometrically consistent initial latent representation that preserves structural attributes. Vulnerability analyses across four state-of-the-art face recognition systems demonstrate that DCMorph achieves the highest attack success rates compared to existing methods at both operational thresholds, while remaining challenging to detect by current morphing attack detection solutions.
Tahar Chettaoui, Eduarda Caldeira, Guray Ozgur +3
Fraunhofer Institute for Computer Graphics Research IGD, Germany · TU Darmstadt, Germany · Norwegian University of Science and Technology (NTNU), Norway
Face morphing attacks pose a substantial risk to the reliability of face recognition systems used in passport issuance, border control, and digital identity verification. Detecting morphing attacks from a single facial image remains challenging owing to the lack of a trusted reference and the diversity of attack generation methods. This paper presents a new Single-Image Face Morphing Attack Detection (S-MAD) framework that integrates high-frequency Laplacian residual statistics with representations from a frozen, foundation-scale vision transformer. The approach employs residual-statistic-gated low-rank adapters (R-FLoRA) and feature-wise residual fusion (Res-FiLM) to enhance sensitivity to local morphing artefacts while preserving the semantic context of the backbone. A novel residual-contrastive alignment loss further regularises the fused token space, improving discrimination under unseen morphing conditions. Comprehensive experiments on four ICAO-compliant datasets, encompassing seven morph generation techniques, demonstrate that the proposed method consistently surpasses nine recent state-of-the-art S-MAD algorithms in detection accuracy and cross-domain (or dataset) generalisation. With a frozen backbone and minimal trainable parameters, the model achieves real-time efficiency and interpretability, making it suitable for real-life scenarios in biometric verification systems.
Raghavendra Ramachandra
SAFE Center, Department of Information Security and Communication Technology, Norwegian University of Science and Technology (NTNU), Norway
Face morphing attacks compromise biometric security by creating document images that verify against multiple identities, posing significant risks from document issuance to border control. Differential Morphing Attack Detection (D-MAD) offers an effective countermeasure, particularly when employing face demorphing to disentangle identities blended in the morph. However, existing methods lack operational generalizability due to limited training data and the assumption that all document inputs are morphs. This paper presents SFDemorpher, a framework designed for the operational deployment of face demorphing for D-MAD that performs identity disentanglement within joint StyleGAN latent and high-dimensional feature spaces. We introduce a dual-pass training strategy handling both morphed and bona fide documents, leveraging a hybrid corpus with predominantly synthetic identities to enhance robustness against unseen distributions. Extensive evaluation confirms state-of-the-art generalizability across unseen identities, diverse capture conditions, and 13 morphing techniques, spanning both border verification and the challenging document enrollment stage. Our framework achieves superior D-MAD performance by widening the margin between the score distributions of bona fide and morphed samples while providing high-fidelity visual reconstructions facilitating explainability.
Raul Ismayilov, Luuk Spreeuwers
Data Management & Biometrics, University of Twente, Drienerlolaan 5, Enschede, 7512AD, Netherlands.