GAMES Webinar 2022 – 260期(当对抗鲁棒性邂逅计算机图形学) | He Wang(the University of Leeds)，Hang Zhou (Simon Fraser University)
【GAMES Webinar 2022-260期】(几何专题-当对抗鲁棒性邂逅计算机图形学)
报告嘉宾：He Wang(the University of Leeds)
报告题目：The Universal Vulnerability of Human Action Recognition Classifiers and Potential Solutions
Deep learning has been regarded as the ‘go to’ solution for many tasks today, but its intrinsic vulnerability to malicious attacks has become a major concern. The vulnerability is affected by a variety of factors including models, tasks, data, and attackers. In this talk, we investigate skeleton-based Human Activity Recognition (S-HAR), which is an important type of time-series data widely used for self-driving cars, security and safety, etc. Very recently, we have identified a universal vulnerability in existing S-HAR classifiers. This is through proposing the first adversarial attack approaches on such tasks/classifiers. Surprisingly, the design of these approaches is highly inspired by the research in computer graphics, especially character animation. Further, we also investigate how to enhance the robustness and resilience of existing classifiers, across different data, tasks, classifiers and attackers.
He Wang is an Associate Professor at the Visualization and Computer Graphics group, in the School of Computing, University of Leeds, UK. He is also a Turing Fellow at the Alan Turing Institute UK, an Academic Advisor at the Commonwealth Scholarship Council, the Director of High-Performance Graphics and Game Engineering, and an academic lead of Centre for Immersive Technology at Leeds. He has over 50 publications in internationally leading conferences (CVPR, ICCV, AAAI, SIGGRAPH, SIGGRAPH Asia, ICRA, etc.) and journals (Science Advances, ACM Transactions on Graphics, IEEE TVCG, etc.) . He leads/co-leads a research portfolio of over £7.3M. His current research interest is mainly in computer graphics, vision and machine learning and applications. Previously he was a Senior Research Associate at Disney Research Los Angeles. He received his PhD from the University of Edinburgh UK and his bachelor’s from Zhejiang University China.
报告嘉宾：Hang Zhou (Simon Fraser University)
报告题目：Exploring Deep Point-Cloud Robustness
Deep point-cloud neural networks such as DeepSets and PointNet are functioning in numerous 3D vision tasks. However, recent works have demonstrated that DNNs are vulnerable to adversarial examples, which would potentially bring security threats to real application systems such as autonomous driving and robotics. In this presentation, I will briefly describe our recent works on generative adversarial attack, shape-invariant adversarial attack, defence by filtering and upsampling, and defence by diffusion pretraining. Finally, I will discuss the challenges and opportunities in this area for future work.
Hang Zhou is a postdoctoral researcher at Simon Fraser University. He received his Ph.D. from University of Science and Technology of China in 2020, and Bachelor’s degree from Shanghai University in 2015. He received Outstanding Doctoral Dissertation from Chinese Academy of Sciences in 2021. His research interest is in Computer Graphics and Machine Learning Robustness.
唐可可，博士，副教授，硕士生导师，广州大学网络空间先进技术研究院（方滨兴院士团队）大数据及安全研究所副所长。本科毕业于吉林大学，博士毕业于中国科学技术大学，之后于香港大学从事博士后研究两年。近年来一直在从事计算机图形学、计算机视觉、机器人以及人工智能安全相关的研究工作，在国际顶级会议和期刊ICCV、AAAI、Eurographics、ICRA、TNNLS等发表论文数十篇，作为中科大“可佳”、“佳佳”机器人核心研发团队成员，两次获得RoboCup机器人世界杯冠军。目前担任CVPR、ICCV、ECCV、AAAI、ACMMM等顶级会议审稿人/程序委员会委员，Security and Communication Networks期刊编委，CCF CAD&CG专委和CSIAM GDC专委委员、GAMES执行委员，主持国家自然科学基金、广东省自然科学基金等纵向课题多项。个人主页：https://tangbohu.github.io/。
GAMES主页的“使用教程”中有 “如何观看GAMES Webinar直播？”及“如何加入GAMES微信群？”的信息；