IAM

Check out our latest research on adversarial robustness and generalization of deep networks.
01stMARCH2019

READING

Aaditya Prakash, Nick Moran, Solomon Garber, Antonella DiLillo, James A. Storer. Protecting JPEG Images Against Adversarial Attacks. DCC, 2018.
Motivated by JPEG compression, Prakash et al. propose an adaptive quantization scheme as defense against adversarial attacks. They argue that JPEG experimentally reduces adversarial noise; however, it is difficult to automatically decide on the level of compression as it also influences a classifier’s performance. Therefore, Prakash et al. use a saliency detector to identify background region, and then apply adaptive quantization – with coarser detail at the background – to reduce the impact of adversarial noise. In experiments, they demonstrate that this approach outperforms simple JPEG compression as defense while having less impact on the image quality.
Also find this summary on ShortScience.org.

What is your opinion on the summarized work? Or do you know related work that is of interest? Let me know your thoughts in the comments below: