Motion aware deepfake videos can be detected with new method
Stop My Dancing! Understanding, Detecting and Attributing Motion-Aware Deepfake Videos
Computer Vision and Pattern RecognitionCryptography and Security
Summary
Fake videos that show moving people can now be made very realistically, which can trick many viewers. The authors created a large test collection of these fake videos and real ones to better understand them. They found these videos have tiny signs in how movement is created that give them away. Using these clues, they built a system called MoDA that can detect and say which method made the fake video with high accuracy. MoDA also works well on unseen fake videos and can still detect fakes even when attackers try to hide them.
What this means in practice
- •For online content moderators: Detect and identify videos using motion-aware deepfake techniques to prevent misinformation and protect platform integrity.
- •For video security teams: Deploy tools to identify sophisticated deepfake videos in forensic or security investigations based on motion artifacts.
Authors
Fazhong Liu, Yan Meng, Tian Dong, Guoxing Chen, Haojin Zhu
Abstract
Pose-guided diffusion models can now synthesize entire human figures in motion, spawning a new class of deepfakes: Motion Aware Deepfake (MAD) that have already reached hundreds of millions of viewers. To better understand this emerging threat, we construct the first MAD-specific benchmark and measurement framework, containing over 1.5 million frames that mix 1,363 real and 30,122 synthetic videos from six controllable generators, with realistic perturbations and open-world evaluation splits. Then, we dissect MAD and discover that, despite their global coherence, these videos betray faint yet reliable cues: because the model relies on limited input frames for motion synthesis, it must predict and simulate coherent movement at motion boundaries, thereby producing high-frequency artifacts along with model-specific spectral fingerprints. Based on the observations obtained from analysis on dataset, we propose MoDA, the first defense framework tailored to detect and attribute MAD videos. MoDA couples spatial semantics with steganalysis-rich frequency features via cross-domain alignment and multi-scale aggregation, achieving 94.8% in-distribution and 89.1% cross-dataset detection accuracy gains of 10% to 25% over prior work and 91.5% model attribution accuracy. MoDA achieves 81.94% accuracy on 200 clips produced by two unseen commercial MAD platforms, indicating promising zero-shot transfer, and 78.13% detection accuracy on 1,200 unseen MAD video clips (55k frames in total) collected from the open Internet. Under white-box, gray-box, and black-box adaptive attacks, MoDA maintains relatively stable detection and attribution performance while the accuracies of the baselines drop rapidly.