all AI news
SSyncOA: Self-synchronizing Object-aligned Watermarking to Resist Cropping-paste Attacks
May 7, 2024, 4:48 a.m. | Chengxin Zhao, Hefei Ling, Sijing Xie, Han Fang, Yaokun Fang, Nan Sun
cs.CV updates on arXiv.org arxiv.org
Abstract: Modern image processing tools have made it easy for attackers to crop the region or object of interest in images and paste it into other images. The challenge this cropping-paste attack poses to the watermarking technology is that it breaks the synchronization of the image watermark, introducing multiple superimposed desynchronization distortions, such as rotation, scaling, and translation. However, current watermarking methods can only resist a single type of desynchronization and cannot be applied to protect …
abstract arxiv attacks challenge cs.cv easy image image processing images modern object processing synchronization technology tools type watermark watermarking
More from arxiv.org / cs.CV updates on arXiv.org
Retrieval-Augmented Egocentric Video Captioning
2 days, 4 hours ago |
arxiv.org
Mirror-Aware Neural Humans
2 days, 4 hours ago |
arxiv.org
Jobs in AI, ML, Big Data
Software Engineer for AI Training Data (School Specific)
@ G2i Inc | Remote
Software Engineer for AI Training Data (Python)
@ G2i Inc | Remote
Software Engineer for AI Training Data (Tier 2)
@ G2i Inc | Remote
Data Engineer
@ Lemon.io | Remote: Europe, LATAM, Canada, UK, Asia, Oceania
Artificial Intelligence – Bioinformatic Expert
@ University of Texas Medical Branch | Galveston, TX
Lead Developer (AI)
@ Cere Network | San Francisco, US