Paired-Sampling Contrastive Framework for Joint Physical-Digital Face Attack Detection
Andrei Balykin, Anvar Ganiev, Denis Kondranin, Kirill Polevoda, Nikolai Liudkevich, Artem Petrov · 2025
Read Paired-Sampling Contrastive Framework for Joint Physical-Digital Face Attack Detection by Andrei Balykin, Anvar Ganiev, Denis Kondranin, Kirill Polevoda, Nikolai Liudkevich, Artem Petrov (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-14980.txt.
Description
Balykin and five co-authors built this face security method in twenty twenty-five. Face recognition can be fooled in two ways: by physical fakes, such as photos or masks, and by digital forgeries. Separate detectors for each kind add cost and leave gaps. The team trained one model to catch both.
Their method pairs each real selfie with a matched attack version during training. Contrasting the pairs teaches the model signs of liveness that work across attack types. On a major challenge benchmark, the method posted a low average error of about two percent. It beat earlier solutions while staying light and quick to train. The team released code and trained models for others to use.
Catalog record
| CREATOR | Andrei Balykin, Anvar Ganiev, Denis Kondranin, Kirill Polevoda, Nikolai Liudkevich, Artem Petrov |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 24 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-14980