Report Title: Adversarial Examples of Deep Learning for Real-World (Physical) Scenarios

Speaker: Professor Zhengyu Zhao

Affiliation: Xi'an Jiaotong University

Time: 15:00-16:00, Wednesday, June 25, 2025

Venue: Room 1104, Block A, Feicui Science and Education Building

Report Abstract:

Artificial intelligence (AI) technologies represented by deep learning are widely used in social, economic and people's livelihood fields. However, while creating enormous value, AI systems deployed in real-world scenarios such as autonomous driving and security monitoring are facing complex and severe security challenges. This report will focus on adversarial examples, a typical AI security threat, and discuss the two key characteristics that adversarial examples should possess in real-world (physical) scenarios: transferability and robustness. In addition, this report will also introduce the attack stealth and defense perspectives.

Speaker Profile:

Zhengyu Zhao is a Professor/PhD Supervisor at the School of Cyberspace Security, Xi'an Jiaotong University, and a National High-level Young Talent. He was a Postdoctoral Researcher at the CISPA Helmholtz Center for Information Security in Germany, and received his Ph.D. from Radboud University in the Netherlands, winning the CVPR Doctoral Forum Award. His main research focuses on artificial intelligence security, and he has published more than 30 papers in international conferences/journals. He serves as an Area Chair of several CCF Class A conferences and a reviewer for conferences/journals, winning 1 Outstanding Area Chair Award and 4 Outstanding Reviewer Awards. He has led the team to win the top 3 in international artificial intelligence security competitions at CCF Class A conferences for many times.