Kaleel Mahmood, On the Robustness of Vision Transformers to Adversarial Examples

When: Friday 2/16 from 10:00 to 11:00 AM
Where: Ranger 202
Abstract: Machine learning has become ubiquitous, being deployed in a range for domains like self driving cars, medical imaging and face recognition. With this increased use of machine learning an important question arises, how secure are these systems? In this presentation we dive into the security vulnerabilities of machine learning. Specifically we analyze the new Vision Transformer architecture to see how robust it is to adversarial attacks and how it can be used to further advance the field of adversarial machine learning.
Bio: Kaleel Mahmood is an Assistant Professor in Residence in the Computer Science department and is an Assistant Research Professor in the Electrical Engineering department at the University of Connecticut (joint appointment). His work focuses on the applications of machine learning to security and vision tasks. He has accepted works in ICCV, IEEE Access and IEEE Transactions on Signal Processing, with over 600 citations to date.