计算机软件新技术国家重点实验室 

摘 要：

Today’s state-of-the-art image classifiers fail to correctly classify carefully manipulated adversarial images. In this talk, we develop a new, localized adversarial attack that generates adversarial examples by imperceptibly altering the backgrounds of normal images.  We then include images with adversarial backgrounds in the training set. This focuses the training on the image foregrounds, increasing accuracy and robustness. Localized adversarial training is cheap to implement and could have broad applications.

报告人简介:

Tingting Chen is an associate professor in Computer Science Department, at California State Polytechnic University, Pomona. She graduated with a Ph.D. degree from Computer Science and Engineering Department, at State University of New York at Buffalo, in June 2011. She received my M.S. degree and B.S. degree both in Computer Science and Engineering from Harbin Institute of Technology, China, in 2006 and 2004 respectively. Her current research interests include wireless networks, data privacy, health informatics and cyber-security. Her research has been supported by National Science Foundation, Amazon Inc, Oklahoma Center for the Advancement of Science and Technology, California State Polytechnic University, and Oklahoma State University.

时间：12月30日  15:00-16:00

地点：计算机科学技术楼229室