报告人：袁星亮 副教授 墨尔本大学
主持人：凌振
报告时间：2025年6月17日（周二）上午9:30-11:30
报告地点：澳门永利集团 九龙湖校区计算机楼513报告厅
报告摘要：As deep learning models increasingly rely on vast datasets, concerns about data privacy, consent, and misuse have grown significantly. Individuals and organisations may discover after the fact that their data has been used in AI training without their authorisation. The "right to be forgotten" has emerged as a critical principle in data protection laws, but enforcing this right in the context of machine learning systems and services remains technically complex. In this talk, I will showcase our initial efforts in detecting and mitigating training data misuse with a focus on graph neural networks (GNNs). I will present how to design privacy-aware membership inference to enable data owners to detect graph data misuse after GNN models are deployed in machine learning as a service, and then leverage synthetic graphs to facilitate machine unlearning, so as to remove the influence of unauthorised training data from deployed models without directly exposing the private unlearning target. Finally, I will make an outlook on remaining challenges and future directions in the broader context of large AI systems.
报告人简介：Dr Xingliang Yuan is an Associate Professor in the School of Computing and Information Systems, the University of Melbourne. Before that, he was a faculty member at Monash University from 2017 to 2024. Xingliang has a keen interest in designing systems to address real-world privacy and security challenges. His research has been supported by Australian Research Council, Commonwealth Scientific and Industrial Research Organisation (CSIRO), Australian Department of Home Affairs, Australian Department of Health and Aged Care, and the Oceania Cyber Security Centre. His work has been published in major venues of computer security and systems, such as ACM CCS, IEEE S&P, USENIX Security, NDSS, TDSC, and TIFS. He is a sole recipient of the Dean's Award for Excellence in Research by an Early Career Researcher (2020), the Faculty Teaching Excellence Award (2021) at Monash, and the Excellence in Engagement award at UniMelb (2024). He is a co-recipient of the best paper award in the European Symposium on Research in Computer Security 2021. He is on the editorial board of IEEE Transactions on Dependable and Secure Computing and IEEE Transactions on Service Computing. He is a general chair of RAID’25, track co-chair of ICDCS'24, and a program co-chair of Lamps@CCS'24, SecTL@AsiaCCS'23, and NSS'22.
