Skip to main content
. 2025 Dec 10;8:1720525. doi: 10.3389/fdata.2025.1720525

Figure 2.

Bar chart comparing the accuracy of different methods in two scenarios: FF++-AF and Celeb-DF-AF. The methods include Xception, FreqNet, SRM, LipForensics, DCNetwork, IID, DFS, Clipping, and Ours. The Ours method shows the highest accuracy in both cases, with slightly higher performance in the Celeb-DF-AF scenario.

Performance comparison on anti-forensic datasets. Our method achieves a significant performance margin over existing detectors.