A comprehensive evaluation framework and efficient regularization strategy for adversarial robustness of deepfake detectors under closed and open set scenariosPublished in Computer Vision and Image Understanding, 2026Direct LinkShare on Twitter Facebook LinkedIn Previous Next