意大利佛罗伦萨大学Alessandro Piva教授来访并作报告

报告人：ALESSANDRO PIVA教授

时间：2026年09月22日  

地点：北京交通大学九教北613

报告主题：Open-World Detection and Attribution of AI-Generated Images

      报告摘要：The rise of generative AI challenges the assumption that photographs and video are reliable evidence of reality. This talk surveys two decades of multimedia forensics research at  the Signal Processing and Communications Lab (LESC), University of Florence  before focusing on recent work: building data infrastructure for the field (DRAGON, a large-scale diffusion-image dataset; an automated realistic-forgery generation pipeline), detecting AI-generated images by fusing low-level forensic residuals with vision-language semantics, and attributing images to their source generator in open-world settings through synthesis-free fingerprinting and lightweight few-shot methods. The unifying requirement throughout: generalization to generators never seen in training.

    报告人简介：ALESSANDRO PIVA is an Associate Professor atthe Department of Information Engineering,University of Florence. He was Assistant Professorduring 2005-2014. His research interests includemultimedia security and protection, digitalwatermarking, multimedia forensics, and image andvideo processing. Prof. Piva is a Fellow of the IEEE.He is the Head of FORLAB and committee memberof the IEEE Information Forensics and SecurityTechnical Committee.He served as Research Scientist at the National Interuniversity Consortium for Telecommunications (CNIT) from 2002 to 2004. Prof.Piva serves as Head of LESC, and is involved in the research activitiesof CNIT and the Media Integration and Communication Center ofExcellence (MICC).