V-FIND: Revealing the Intrinsic Forgery Knowledge Encoded in Video Forgery Detectors
Quick Answer
This paper shows that The V-FIND framework uncovers and activates sparse forensic knowledge in video forgery detectors, enhancing detection performance without full model retraining.
Quick Take
By identifying specialized neurons, it organizes them into a compact subspace, achieving strong results across benchmarks while maintaining the original model's integrity.
Key Points
- V-FIND identifies critical layers with discrepancies between real and forged videos.
- The framework organizes specialized neurons into a compact forensic subspace.
- Strong detection performance achieved with a lightweight linear classifier.
- Neuron intervention experiments confirm the functional specificity of discovered neurons.
- The study suggests extractable and reusable knowledge within video forgery detectors.
Paper Resources
📖 Reader Mode
~2 min readAuthors:Shichao Kan, Chengpeng Hong, Jingtong Dou, Chuancheng Shi, Yuhan Liu, Linrui Xu, Yixiong Liang, Yigang Cen, Yanpeng Sun, Fei Shen, Tat-Seng Chua
Abstract:As generated videos become increasingly realistic, reliable video forgery detection is increasingly important. Existing studies typically optimize and use video forgery detectors as black boxes, while the latent forgery-discriminative knowledge inside them remains largely unexplored. Instead of continuing to rely on resource-intensive full-model retraining to steadily improve detection performance, we ask whether video forgery detection can also be achieved by uncovering and activating sparse forensic knowledge within the detector. We find that forgery-discriminative knowledge is not uniformly distributed across the full representation space, but is concentrated in a sparse set of functionally specialized neurons. Based on this insight, we propose a video forgery-intrinsic neuron discovery (V-FIND) framework. V-FIND first localizes critical layers that exhibit pronounced discrepancies between real and forged videos, and then identifies latent anchor neurons that consistently carry forgery-discriminative signals, organizing them into a compact forensic subspace. With the original backbone frozen and only a lightweight linear classifier trained, this subspace still delivers strong detection performance across multiple external benchmarks for generated videos. Further neuron intervention experiments provide direct evidence for the functional specificity of the discovered neurons. Overall, these results suggest that video forgery detectors contain sparse, extractable, and reusable forgery-discriminative knowledge, offering a new perspective on understanding and exploiting their intrinsic forensic capability.
| Comments: | 12 pages, 12 figures. Under review |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.03008 [cs.CV] |
| (or arXiv:2608.03008v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2608.03008 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Chengpeng Hong [view email]
[v1]
Tue, 4 Aug 2026 01:41:40 UTC (2,078 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing Modalities
ProMoE-FL introduces a Prototype-conditioned Mixture-of-Experts framework for multimodal federated learning, effectively addressing missing modalities. It outperforms existing methods on four chest X-ray datasets, demonstrating superior feature synthesis capabilities in both homogeneous and heterogeneous settings.