Papers for

digital media forensics teams

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

AI detects fake videos using text cues and timing patterns

MTOR: Generalizable AI-Generated Video Detection with Multimodal Semantics and Temporal Over-Regularity

Abstract: The rapid evolution of video generation has narrowed the perceptual gap between authentic and synthetic videos, making generalizable AI-generated video detection increasingly challenging. Existing detectors predominantly rely on visual representations, leaving caption-derived textual semantics underexplored. Meanwhile, temporal regularity in fine-grained visual representations has received limited attention. We find that caption-derived textual representations provide complementary discriminative cues to global visual representations. Our analysis further reveals that AI-generated videos exhibit stronger temporal persistence and lower temporal variability, a pattern we term temporal over-regularity (TOR). Based on these findings, we propose MTOR with a multimodal branch and a TOR component. The multimodal branch integrates global visual and caption-derived textual representations, while the TOR component models temporal over-regularity at three levels: coarse inter-frame continuity, fine-grained token correspondence, and frame-to-video stability. Extensive evaluations on five benchmarks covering 46 generator variants demonstrate state-of-the-art overall performance against 16 representative baselines, while robustness experiments confirm strong resilience to twelve real-world video perturbations. Code and models will be released at https://github.com/hwang-cs-ime/MTOR.

Mon 5 OctComputer Vision and Pattern Recognition

The gist

It is getting harder to tell real videos from AI-made fake ones using traditional methods that look only at images. The authors found that looking at the text captions describing the videos gives helpful clues to spot fakes. They also noticed fake videos tend to be more regular and less varied over time in certain ways, which they call temporal over-regularity. Combining these insights, the authors built a new detection method called MTOR that looks at both video images and captions, and checks how video content changes over time. Their tests show MTOR works better than many existing methods, and it stays reliable even when videos are altered in common ways.

Open → 2610.06378v1

Ensemble method improves model identification for synthetic images

A Multi-View and Confusion-Guided Ensemble Framework for Robust Synthetic Image Attribution

Abstract: Synthetic image attribution (SIA) has become increasingly important with the rapid advancement of text-to-image generation models. However, accurately identifying the source model of a generated image remains challenging due to the growing similarity among modern diffusion-based generators and the presence of diverse post-processing operations. In this report, we present a multi-view and confusion-guided ensemble framework for the Synthetic Image Attribution Challenge of the DLMMDD Workshop at ICANN 2026. Our approach integrates multiple complementary architectures, including FFT-ConvNeXt, DINOv2, CLIP, and Xception, to capture diverse attribution cues from frequency, semantic, and forensic perspectives. To improve robustness against unknown degradations and image manipulations, extensive data augmentation strategies are employed during training, simulating realistic post-processing operations such as compression, resizing, grayscale conversion, and blur. Furthermore, we analyze the confusion patterns of the ensemble model and observe severe ambiguity between Stable Diffusion 3 and Stable Diffusion 3.5. To address this issue, we introduce a dedicated binary expert classifier that is selectively activated under low-confidence conditions. We additionally apply class-adaptive confidence calibration to improve the discrimination of challenging classes such as Tencent Hunyuan. The proposed framework achieved 99.53% on the public leaderboard and 99.20% on the private leaderboard. The source code and implementation details are publicly available at https://github.com/ZOMIN28/SIA.

Thu 10 SeptComputer Vision and Pattern Recognition

The gist

Identifying which AI model created a synthetic image is hard because many modern image generators produce very similar images and these images can be edited afterward. The authors combined several different techniques that look at images in multiple ways, such as their frequency patterns and semantic content, to catch unique clues. They also trained their system to handle common changes like resizing or blurring. When their system was unsure, it used a special focused classifier to tell closely related image models apart. This approach worked really well in a competition, achieving over 99% accuracy.

Open → 2609.11188v1