Deep Learning Seminar  /  December 10, 2020, 10:00 – 11:00

Adversarial Examples – An Introduction

Speaker: Paula Harder (Fraunhofer ITWM, department HPC)

Abstract:

Despite the success of Deep Neural Networks (DNNs) in multiple domains, they remain vulnerable against so called adversarial examples. A small, crafted perturbation in the input image can lead to missclassification. In this talk an introduction to the concepts and ideas of those attacks will be given, as well as defense possibilities and further application cases.