A Dataset for Assessing the Stability of Fake Video Detection in Video Conferencing
Programmnaya Ingeneria · 10.08.2026 · 17 (8) · 426-434
DOI: 10.17587/prin.17.426-434 ↗Kurz gefasst
Fachartikel in der Zeitschrift Programmnaya Ingeneria, erschienen am 10.08.2026. Verfasst von M. D. Krasilnikov und 2 weiteren Autor:innen.
Abstract (Originalsprache)
The FakeMeet dataset is proposed for evaluating the robustness of face-swapping detection algorithms in video recordings under video-conferencing conditions. FakeMeet recreates typical video-call conditions by modeling distortions caused by resolution changes and by H.264 encoding with different compression levels, and it also includes variation in the background. An experimental comparison of fourteen modern detectors is conducted in a cross-domain evaluation setting on the Celeb-DF and FakeMeet datasets. It is shown that, under conditions characteristic of video conferencing, detection performance noticeably decreases. These findings point to the need for stable approaches that remain reliable under resolution shifts, compression artifacts, and the use of different generation methods. FakeMeet provides a standardized, scenario-specific benchmark to drive advancements in securing digital communication platforms against evolving deepfake threats.
Zitieren
APA
Krasilnikov, M. D., Nikitin, M. Y., & Konushin, A. S. (2026). A Dataset for Assessing the Stability of Fake Video Detection in Video Conferencing. Programmnaya Ingeneria, 17(8), 426-434. https://doi.org/10.17587/prin.17.426-434
Vancouver
Krasilnikov MD, Nikitin MY, Konushin AS. A Dataset for Assessing the Stability of Fake Video Detection in Video Conferencing. Programmnaya Ingeneria. 2026;17(8):426-434. doi:10.17587/prin.17.426-434
BibTeX
@article{Krasilnikov2026,
author = {Krasilnikov, M. D. and Nikitin, M. Yu. and Konushin, A. S.},
title = {A Dataset for Assessing the Stability of Fake Video Detection in Video Conferencing},
journal = {Programmnaya Ingeneria},
year = {2026},
volume = {17},
number = {8},
pages = {426--434},
doi = {10.17587/prin.17.426-434}
}
EUVIMED