Metraipolis › archive › dc_8a51d543005f
Damnatio Memoriae: Adversarially and Selectively Forgetting Identities in the Embedding Space of Face Recognition Models — arXiv:2609.27115
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.27115
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T12:34:39.563Z
- SHA-256 of the archived text
6d2cbb0ab7621cda9a9578d98c6951fe0509c649b4e56cb90467080ccd243b55- Stable URL of this page
https://metraipolis.com/documento/dc_8a51d543005f
How to cite
Damnatio Memoriae: Adversarially and Selectively Forgetting Identities in the Embedding Space of Face Recognition Models — arXiv:2609.27115 — arXiv (en), https://arxiv.org/abs/2609.27115 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8a51d543005f (Metraipolis archive dc_8a51d543005f, sha256 6d2cbb0ab7621cda9a9578d98c6951fe0509c649b4e56cb90467080ccd243b55)
BibTeX
@misc{dc8a51d543005f,
title = {Damnatio Memoriae: Adversarially and Selectively Forgetting Identities in the Embedding Space of Face Recognition Models — arXiv:2609.27115},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.27115},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8a51d543005f; sha256 6d2cbb0ab7621cda9a9578d98c6951fe0509c649b4e56cb90467080ccd243b55},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.27115. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
A face recognition model links two images of a person recorded on separate occasions when their embedding similarity exceeds an operating threshold. We consider making chosen identities unlinkable across separate occasions while the model remains in service for the rest of the population. Deleting their images and retraining does not achieve this, since the model recognises identities never observed in training. Therefore, the embedding space must be altered against these identities, the process of which we call open-set adversarial forgetting. We propose three loss functions, one that disperses an identity's embeddings from their centroid, and two that map each image onto its own near-orthogonal target, learnt with the classifier head or fixed in advance as an almost-orthonormal frame. Each is fine-tuned alongside the classification objective on a subset of each identity's images. We evaluate them against four methods from prior work in verification and identification, at two forget scales and three backbones. Every loss acting on the embedding geometry makes the forget identities nearly unidentifiable. The orthonormal frame alone achieves strong forgetting, which holds wherever an image of that subset enters the comparison and leaves distinct forget identities unlinkable. It also surpasses a concurrent unsupervised method at a higher retain rate. Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-24 in cs.CV). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2609.27115 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-08T12:12:35.641Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-24?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).