![[2602.13168] Realistic Face Reconstruction from Facial Embeddings via Diffusion Models](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.13168] Realistic Face Reconstruction from Facial Embeddings via Diffusion Models
Summary
This paper presents a novel framework for reconstructing realistic high-resolution face images from facial embeddings using diffusion models, addressing privacy concerns in face recognition systems.
Why It Matters
As face recognition technology advances, ensuring privacy is crucial. This study highlights potential vulnerabilities in privacy-preserving face recognition systems by demonstrating how realistic faces can be reconstructed from embeddings, thus raising awareness about privacy risks and the need for improved security measures.
Key Takeaways
- Introduces the face embedding mapping (FEM) framework for face reconstruction.
- Demonstrates the ability to reconstruct faces from partial and protected embeddings.
- Validates the method's effectiveness against state-of-the-art face recognition systems.
- Highlights the implications for privacy risks in face recognition technologies.
- Suggests FEM as a tool for evaluating the safety of face recognition systems.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.13168 (cs) [Submitted on 13 Feb 2026] Title:Realistic Face Reconstruction from Facial Embeddings via Diffusion Models Authors:Dong Han, Yong Li, Joachim Denzler View a PDF of the paper titled Realistic Face Reconstruction from Facial Embeddings via Diffusion Models, by Dong Han and 2 other authors View PDF HTML (experimental) Abstract:With the advancement of face recognition (FR) systems, privacy-preserving face recognition (PPFR) systems have gained popularity for their accurate recognition, enhanced facial privacy protection, and robustness to various attacks. However, there are limited studies to further verify privacy risks by reconstructing realistic high-resolution face images from embeddings of these systems, especially for PPFR. In this work, we propose the face embedding mapping (FEM), a general framework that explores Kolmogorov-Arnold Network (KAN) for conducting the embedding-to-face attack by leveraging pre-trained Identity-Preserving diffusion model against state-of-the-art (SOTA) FR and PPFR systems. Based on extensive experiments, we verify that reconstructed faces can be used for accessing other real-word FR systems. Besides, the proposed method shows the robustness in reconstructing faces from the partial and protected face embeddings. Moreover, FEM can be utilized as a tool for evaluating safety of FR and PPFR systems in terms of privacy leakage. All images used in this work are from p...
