VISION: a video and image dataset for source identification

摘要: Forensic research community keeps proposing new techniques to analyze digital images and videos. However, the performance of proposed tools are usually tested on data that far from reality in terms resolution, source device, processing history. Remarkably, latest years, portable devices became preferred means capture videos, contents commonly shared through social media platforms (SMPs, for example, Facebook, YouTube, etc.). These facts pose challenges forensic community: most modern cameras feature stabilization, is proved severely hinder video identification technologies; moreover, strong re-compression enforced by SMPs during upload threatens reliability multimedia tools. On other hand, both videos with same sensor, opening opportunities. The goal this paper propose VISION dataset as a contribution development forensics. currently composed 34,427 1914 native format their version (Facebook, WhatsApp considered), 35 11 major brands. can be exploited benchmark exhaustive evaluation several image