Fetching the paper…

Reading the bibliography…

2016

A study of the effect of JPG compression on adversarial images

Dziugaite, Gintare Karolina, Ghahramani, Zoubin, Roy, Daniel M.

Original
PDF
Post on X

Understand

Neural network image classifiers are known to be vulnerable to adversarial images, i.e., natural images which have been modified by an adversarial perturbation specifically designed to be imperceptible to humans yet fool the classifier.

Not only can adversarial images be generated easily, but these images will often be adversarial for networks trained on disjoint subsets of data or with different architectures.
Adversarial images represent a potential security risk as well as a serious machine learning challenge---it is clear that vulnerable neural networks perceive images very differently from humans.
Noting that virtually every image classification data set is composed of JPG images, we evaluate the effect of JPG compression on the classification of adversarial images.

A study of the effect of JPG compression on adversarial images · Around

Built on

“Bayesian Learning via Stochastic Gradient Langevin Dynamics”
Original
M. Welling and Y.˜W. Teh · 2011
Earlier work this paper cites.
“Intriguing properties of neural networks”, 2013
Original
Christian Szegedy, Wojciech Zaremba, Ilya Sutskever, Joan Bruna, Dumitru Erhan, Ian˜J. Goodfellow and Rob Fergus · 2013
Earlier work this paper cites.
“Towards Deep Neural Network Architectures Robust to Adversarial Examples”, 2014
Original
Shixiang Gu and Luca Rigazio · 2014
Earlier work this paper cites.
“Explaining and Harnessing Adversarial Examples”, 2014
Original
Ian˜J. Goodfellow, Jonathon Shlens and Christian Szegedy · 2014
Earlier work this paper cites.
“Analysis of classifiers’ robustness to adversarial perturbations”, 2015
Original
Alhussein Fawzi, Omar Fawzi and Pascal Frossard · 2015
Earlier work this paper cites.

Similar

“Deep Residual Learning for Image Recognition”, 2015
Original
Kaiming He, Xiangyu Zhang, Shaoqing Ren and Jian Sun · 2015
Cited alongside, same era.
“Preconditioned Stochastic Gradient Langevin Dynamics for Deep Neural Networks”, 2015
Original
Chunyuan Li, Changyou Chen, David Carlson and Lawrence Carin · 2015
Cited alongside, same era.
“Distillation as a Defense to Adversarial Perturbations against Deep Neural Networks”, 2015
Original
Nicolas Papernot, Patrick˜Drew McDaniel, Xi Wu, Somesh Jha and Ananthram Swami · 2015
Cited alongside, same era.
“ImageNet Large Scale Visual Recognition Challenge”
Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander˜C. Berg and Li Fei-Fei · 2015
Cited alongside, same era.
“Rethinking the Inception Architecture for Computer Vision”, 2015

Then

“Exploring the Space of Adversarial Images”, 2015
Original
Pedro Tabacof and Eduardo Valle · 2015
Later among the works it cites.
“Hitting Depth: Investigating Robustness to Adversarial Examples in Deep Convolutional Neural Networks” http://cs231n.stanford.edu/reports2016/119_Report.pdf , 2016
Chris Billovits, Mihail Eric and Nipun Agarwala · 2016
Closest in time.
“Adversarial examples in the physical world”, 2016
Original
Alexey Kurakin, Ian Goodfellow and Samy Bengio · 2016
Closest in time.
“Practical Black-Box Attacks against Deep Learning Systems using Adversarial Examples”, 2016
Original
Nicolas Papernot, Patrick˜Drew McDaniel, Ian˜J. Goodfellow, Somesh Jha, Z.˜Berkay Celik and Ananthram Swami · 2016
Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…

Original

Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens and Zbigniew Wojna · 2015

Cited alongside, same era.