Short bio: Yong Xiang received the Ph.D. degree in Electrical and Electronic Engineering from The University of Melbourne, Australia. He is a Professor at the School of Information Technology, Deakin University, Australia. His research interests include cybersecurity, machine learning & AI, distributed computing, and communications engineering. He has published 7 authored books, over 310 refereed journal articles, and over 120 conference papers in these areas, most of which were published in prestigious venues. His outputs have attracted over 20,400 citations with an h-index of 70 (Google Scholar). Professor Xiang has secured over $7.8 million research funds, including 5 Discovery Projects (4 as Lead CI) and 4 Linkage Projects (3 as Lead CI) from the Australian Research Council, and 19 projects from industry. Professor Xiang is the Associate Editor of IEEE Communications Surveys and Tutorials, IEEE Transactions on Network Science and Engineering, IEEE Transactions on Cloud Computing, and Computer Standards and Interfaces. He was the Senior Area Editor of IEEE Signal Processing Letters and the Guest Editor of several IEEE journals. He has served as Honorary Chair, General Chair, Program Chair, TPC Chair, Symposium Chair, Track Chair, Publicity Chair, and Session Chair for many conferences, and was invited to give keynotes at numerous conferences.
Abstract: Image watermarking has been widely used to protect the ownership of digital images. Traditionally, this was accomplished using signal-processing techniques; however, such methods are vulnerable to image-processing and forgery attacks. While the rise of deep learning–based watermarking has substantially improved robustness against a wide range of image-processing attacks, it still remains susceptible to some forgery attacks. Such attacks can hijack or transfer ownership, as well as remove the watermark, without altering the semantic content of the image. This talk focuses on deep learning-based image watermarking against forgery attacks. First, I will provide an overview of the background of image watermarking, including the evolution from traditional signal-processing techniques to modern deep learning–based methods. Second, I will discuss how the current deep learning–based image watermarking methods tackle forgery attacks. Finally, I will highlight the open research challenges in this area and outline potential directions for addressing forgery attacks.