Frequency-Aware Robustness Analysis of Deepfake Detection Models
Journal on Artificial Intelligence · 2026 · 8 (1) · 153-167
DOI: 10.32604/jai.2026.078014 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal on Artificial Intelligence, 2026. Verfasst von Haoyang Xu.
Abstract (Originalsprache)
: This paper conducted a comprehensive study on the robustness of three widely used DFD deep learning models—namely, ResNet50, FreqNet, and Xception v1—to controlled perturbation attacks and frequency masking across a range of 12 different distortions. The study was performed on 254,166 ForenSynth test images, characterizing the distribution of FSI-drop values derived from over 3.05 million paired predictions. The distribution of FSI-drop values is sharply peaked around zero: 99.7% of the samples exhibit ∣ Δ ∣ 2 to > 16 masking scale), Xception exhibits moderate attenuation, whereas ResNet50 remains statistically unchanged (median ∣ Δ ∣ < 10 − 4 ). All of these differences are statistically significant at the model level ( p < 0.001). The experiments offer a concrete demonstration that CNNs effectively retain prediction invariance to small amounts of image deformation, but the frequency-sensitive design demonstrates readily interpretable high-frequency sensitivity, thereby providing a principled framework for designing detectors robust to image perturbations. It should be noted that FSI-drop measures score stability rather than absolute performance; §4.6 discusses the complementary AUC curves and leaves combined-distortion or adversarial stress tests to future work.
Zitieren
APA
Xu, H. (2026). Frequency-Aware Robustness Analysis of Deepfake Detection Models. Journal on Artificial Intelligence, 8(1), 153-167. https://doi.org/10.32604/jai.2026.078014
Vancouver
Xu H. Frequency-Aware Robustness Analysis of Deepfake Detection Models. Journal on Artificial Intelligence. 2026;8(1):153-167. doi:10.32604/jai.2026.078014
BibTeX
@article{Xu2026,
author = {Xu, Haoyang},
title = {Frequency-Aware Robustness Analysis of Deepfake Detection Models},
journal = {Journal on Artificial Intelligence},
year = {2026},
volume = {8},
number = {1},
pages = {153--167},
doi = {10.32604/jai.2026.078014}
}
EUVIMED