Connecting the Dots: Detecting Adversarial Perturbations Using Context Inconsistency
Abstract
There has been a recent surge in research on adversarial perturbations that defeat Deep Neural Networks (DNNs); most of these attacks target object classifiers. Inspired by the observation that humans are able to recognize objects that appear out of place in a scene or along with other unlikely objects, we augment the DNN with a system that learns context consistency rules during training and checks for the violations of the same during testing. In brief, our approach builds a set of autoencoders, one for each object class, appropriately trained so as to output a discrepancy between the input and output if a perturbation was added to the sample and trigger context violation. Experiments on PASCAL VOC and MS COCO show that our method effectively detects various adversarial attacks and achieves high ROC-AUC (over 0.95 in most cases); this corresponds to over 20-45 % improvement over a baseline context agnostic method.
Cite
Text
Li et al. "Connecting the Dots: Detecting Adversarial Perturbations Using Context Inconsistency." Proceedings of the European Conference on Computer Vision (ECCV), 2020. doi:10.1007/978-3-030-58592-1_24Markdown
[Li et al. "Connecting the Dots: Detecting Adversarial Perturbations Using Context Inconsistency." Proceedings of the European Conference on Computer Vision (ECCV), 2020.](https://mlanthology.org/eccv/2020/li2020eccv-connecting/) doi:10.1007/978-3-030-58592-1_24BibTeX
@inproceedings{li2020eccv-connecting,
title = {{Connecting the Dots: Detecting Adversarial Perturbations Using Context Inconsistency}},
author = {Li, Shasha and Zhu, Shitong and Paul, Sudipta and Roy-Chowdhury, Amit and Song, Chengyu and Krishnamurthy, Srikanth and Swami, Ananthram and Chan, Kevin S},
booktitle = {Proceedings of the European Conference on Computer Vision (ECCV)},
year = {2020},
doi = {10.1007/978-3-030-58592-1_24},
url = {https://mlanthology.org/eccv/2020/li2020eccv-connecting/}
}