defense arXiv Nov 12, 2025 · Nov 2025
Yanlin Wu, Xiaogang Yuan, Dezhi An · Gansu University of Political Science and Law
Detects AI-generated videos by analyzing latent-space initial noise patterns via diffusion model inversion with cross-generator generalization
Output Integrity Attack visiongenerative
AI-generated video has advanced rapidly and poses serious challenges to content security and forensic analysis. Existing detectors rely mainly on pixel-level visual cues and generalize poorly to unseen generators. We propose DBINDS, a diffusion-model-inversion based detector that analyzes latent-space dynamics rather than pixels. We find that initial noise sequences recovered by diffusion inversion differ systematically between real and generated videos. Building on this, DBINDS forms an Initial Noise Difference Sequence (INDS) and extracts multi-domain, multi-scale features. With feature optimization and a LightGBM classifier tuned by Bayesian search, DBINDS (trained on a single generator) achieves strong cross-generator performance on GenVidBench, demonstrating good generalization and robustness in limited-data settings.
diffusion traditional_ml Gansu University of Political Science and Law
ML Security Papers