×

Adversarial Imaging

17 January. 2021
Adversarial imaging is a term used to identify strategy to subtly modifying images to make image-based deep-learning system to fail (adversarial attack). This work would like to investigate several strategies that can be used to reproduce adversarial images. Furthermore, statistical analysis applied on these images, will be employed to understand how to extend existing image based deep-learning systems to be able to identify these adversarial attacks.
Required Skills
Programming skills in C/C++, Python, knowledge of deep-learning is a plus
Skills Level
Good