ME-Net: Towards effective adversarial robustness with matrix estimation
Copyright © 2019 ASME Deep neural networks are vulnerable to adversarial attacks. The literature is rich with algorithms that can easily craft successful adversarial examples. In contrast, the performance of defense techniques still lags behind. This paper proposes ME-Net, a defense method that leve...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
2021
|
Online Access: | https://hdl.handle.net/1721.1/132310 |
_version_ | 1811083935216566272 |
---|---|
author | Yang, Y Zhang, G Katabi, D Xu, Z |
author_facet | Yang, Y Zhang, G Katabi, D Xu, Z |
author_sort | Yang, Y |
collection | MIT |
description | Copyright © 2019 ASME Deep neural networks are vulnerable to adversarial attacks. The literature is rich with algorithms that can easily craft successful adversarial examples. In contrast, the performance of defense techniques still lags behind. This paper proposes ME-Net, a defense method that leverages matrix estimation (ME). In ME-Net, images are preprocessed using two steps: first pixels are randomly dropped from the image; then, the image is reconstructed using ME. We show that this process destroys the adversarial structure of the noise, while rc-enforcing the global structure in the original image. Since humans typically rely on such global structures in classifying images, the process makes the network mode compatible with human perception. We conduct comprehensive experiments on prevailing benchmarks such as MNIST, CIFAR-10, SVHN, and Tiny-ImageNet. Comparing ME-Net with state-of-the-art defense mechanisms shows that ME-Net consistently outperforms prior techniques, improving robustness against both black-box and white-box attacks. |
first_indexed | 2024-09-23T12:41:57Z |
format | Article |
id | mit-1721.1/132310 |
institution | Massachusetts Institute of Technology |
language | English |
last_indexed | 2024-09-23T12:41:57Z |
publishDate | 2021 |
record_format | dspace |
spelling | mit-1721.1/1323102021-09-21T03:34:33Z ME-Net: Towards effective adversarial robustness with matrix estimation Yang, Y Zhang, G Katabi, D Xu, Z Copyright © 2019 ASME Deep neural networks are vulnerable to adversarial attacks. The literature is rich with algorithms that can easily craft successful adversarial examples. In contrast, the performance of defense techniques still lags behind. This paper proposes ME-Net, a defense method that leverages matrix estimation (ME). In ME-Net, images are preprocessed using two steps: first pixels are randomly dropped from the image; then, the image is reconstructed using ME. We show that this process destroys the adversarial structure of the noise, while rc-enforcing the global structure in the original image. Since humans typically rely on such global structures in classifying images, the process makes the network mode compatible with human perception. We conduct comprehensive experiments on prevailing benchmarks such as MNIST, CIFAR-10, SVHN, and Tiny-ImageNet. Comparing ME-Net with state-of-the-art defense mechanisms shows that ME-Net consistently outperforms prior techniques, improving robustness against both black-box and white-box attacks. 2021-09-20T18:21:47Z 2021-09-20T18:21:47Z 2020-12-23T16:21:37Z Article http://purl.org/eprint/type/ConferencePaper https://hdl.handle.net/1721.1/132310 en http://proceedings.mlr.press/v97/yang19e.html 36th International Conference on Machine Learning, ICML 2019 Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf arXiv |
spellingShingle | Yang, Y Zhang, G Katabi, D Xu, Z ME-Net: Towards effective adversarial robustness with matrix estimation |
title | ME-Net: Towards effective adversarial robustness with matrix estimation |
title_full | ME-Net: Towards effective adversarial robustness with matrix estimation |
title_fullStr | ME-Net: Towards effective adversarial robustness with matrix estimation |
title_full_unstemmed | ME-Net: Towards effective adversarial robustness with matrix estimation |
title_short | ME-Net: Towards effective adversarial robustness with matrix estimation |
title_sort | me net towards effective adversarial robustness with matrix estimation |
url | https://hdl.handle.net/1721.1/132310 |
work_keys_str_mv | AT yangy menettowardseffectiveadversarialrobustnesswithmatrixestimation AT zhangg menettowardseffectiveadversarialrobustnesswithmatrixestimation AT katabid menettowardseffectiveadversarialrobustnesswithmatrixestimation AT xuz menettowardseffectiveadversarialrobustnesswithmatrixestimation |