Filters








172,110 Hits in 2.9 sec

Generating Natural Adversarial Examples [article]

Zhengli Zhao, Dheeru Dua, Sameer Singh
2018 arXiv   pre-print
In this paper, we propose a framework to generate natural and legible adversarial examples that lie on the data manifold, by searching in semantic space of dense and continuous data representation, utilizing  ...  We include experiments to show that the generated adversaries are natural, legible to humans, and useful in evaluating and analyzing black-box classifiers.  ...  FRAMEWORK FOR GENERATING NATURAL ADVERSARIES In this section, we describe the problem setup and details of our framework for generating natural adversarial examples of both continuous images and discrete  ... 
arXiv:1710.11342v2 fatcat:urv6jnxcgfetpfloq7gjdowdtm

Generating Natural Language Adversarial Examples [article]

Moustafa Alzantot, Yash Sharma, Ahmed Elgohary, Bo-Jhang Ho, Mani Srivastava, Kai-Wei Chang
2018 arXiv   pre-print
Given these challenges, we use a black-box population-based optimization algorithm to generate semantically and syntactically similar adversarial examples that fool well-trained sentiment analysis and  ...  Deep neural networks (DNNs) are vulnerable to adversarial examples, perturbations to correctly classified examples which can cause the model to misclassify.  ...  Relative to the image domain, little work has been pursued for generating natural language adversarial examples.  ... 
arXiv:1804.07998v2 fatcat:c5kknmwlprfwdotgnswgkdjdby

Generating Natural Language Adversarial Examples

Moustafa Alzantot, Yash Sharma, Ahmed Elgohary, Bo-Jhang Ho, Mani Srivastava, Kai-Wei Chang
2018 Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing  
Given these challenges, we use a black-box population-based optimization algorithm to generate semantically and syntactically similar adversarial examples that fool well-trained sentiment analysis and  ...  Deep neural networks (DNNs) are vulnerable to adversarial examples, perturbations to correctly classified examples which can cause the model to misclassify.  ...  Relative to the image domain, little work has been pursued for generating natural language adversarial examples.  ... 
doi:10.18653/v1/d18-1316 dblp:conf/emnlp/AlzantotSEHSC18 fatcat:orn7hr6hprhptpecx46q2oedqa

Generating Fluent Adversarial Examples for Natural Languages [article]

Huangzhao Zhang, Hao Zhou, Ning Miao, Lei Li
2020 arXiv   pre-print
Secondly, the fluency of the generated examples cannot be guaranteed.  ...  Efficiently building an adversarial attacker for natural language processing (NLP) tasks is a real challenge.  ...  Conclusion In this paper, we propose MHA, which generates adversarial examples for natural language by adopting the MH sampling approach.  ... 
arXiv:2007.06174v1 fatcat:kvl555rxpfd7np62fid3bqyuqe

Generating Fluent Adversarial Examples for Natural Languages

Huangzhao Zhang, Hao Zhou, Ning Miao, Lei Li
2019 Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics  
Secondly, the fluency of the generated examples cannot be guaranteed.  ...  Efficiently building an adversarial attacker for natural language processing (NLP) tasks is a real challenge.  ...  Conclusion In this paper, we propose MHA, which generates adversarial examples for natural language by adopting the MH sampling approach.  ... 
doi:10.18653/v1/p19-1559 dblp:conf/acl/ZhangZML19 fatcat:ersm2mvcqnh5hf2x5yfuft3esu

A Geometry-Inspired Attack for Generating Natural Language Adversarial Examples [article]

Zhao Meng, Roger Wattenhofer
2020 arXiv   pre-print
Generating adversarial examples for natural language is hard, as natural language consists of discrete symbols, and examples are often of variable lengths.  ...  In this paper, we propose a geometry-inspired attack for generating natural language adversarial examples.  ...  Generating adversarial examples for natural language is fundamentally different from generating adversarial examples for images.  ... 
arXiv:2010.01345v1 fatcat:f7mp4aqztvbsbmv6r4p32tcjwm

Generating Natural Adversarial Hyperspectral examples with a modified Wasserstein GAN [article]

Jean-Christophe Burnel, Nicolas Courty
2020 arXiv   pre-print
In this paper, we present a new method which is able to generate natural adversarial examples from the true data following the second paradigm.  ...  Based on Generative Adversarial Networks (GANs) [5], it reweights the true data empirical distribution to encourage the classifier to generate ad-versarial examples.  ...  So the adversarial examples look like true images and are adversarial for the attacked classifier. We propose a different approach to generate natural adversarial examples from those methods.  ... 
arXiv:2001.09993v1 fatcat:3d2qzur3vrgu3muemmxi5jltru

Generating Natural Language Adversarial Examples through Probability Weighted Word Saliency

Shuhuai Ren, Yihe Deng, Kun He, Wanxiang Che
2019 Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics  
The challenge of this task is to generate adversarial examples that maintain lexical correctness, grammatical correctness and semantic similarity.  ...  A human evaluation study shows that our generated adversarial examples maintain the semantic similarity well and are hard for humans to perceive.  ...  In the following, we first give a definition of adversarial example for natural language classification, and then introduce our word substitution strategy.  ... 
doi:10.18653/v1/p19-1103 dblp:conf/acl/RenDHC19 fatcat:cgbeqhguzjh7vets2af7vjzwjm

Generating Natural Language Adversarial Examples through An Improved Beam Search Algorithm [article]

Tengfei Zhao, Zhaocheng Ge, Hanping Hu, Dingmeng Shi
2021 arXiv   pre-print
Also, further experiments show the novel method has a good transferability on the generated adversarial examples.  ...  However, these attack methods are inefficient as they require lots of queries for the victim model when crafting text adversarial examples.  ...  adversarial examples for tasks such as Sentiment Analysis and Natural Lan- guage Inference. 2.  ... 
arXiv:2110.08036v1 fatcat:zipry7m2enfg3jlxzw3cg5kyge

Generating Natural Language Adversarial Examples on a Large Scale with Generative Models [article]

Yankun Ren and Jianbin Lin and Siliang Tang and Jun Zhou and Shuang Yang and Yuan Qi and Xiang Ren
2020 arXiv   pre-print
Specifically, we train a conditional variational autoencoder (VAE) with an additional adversarial loss to guide the generation of adversarial examples.  ...  However, these classifiers are found to be easily fooled by adversarial examples.  ...  In natural language texts, even a single word change may change the whole meaning of a sentence. A valid adversarial example must be imperceptible to humans.  ... 
arXiv:2003.10388v1 fatcat:tuwfajw7nzblhgeiu72zjxxogq

A Geometry-Inspired Attack for Generating Natural Language Adversarial Examples

Zhao Meng, Roger Wattenhofer
2020 Proceedings of the 28th International Conference on Computational Linguistics   unpublished
Generating adversarial examples for natural language is hard, as natural language consists of discrete symbols, and examples are often of variable lengths.  ...  In this paper, we propose a geometryinspired attack for generating natural language adversarial examples.  ...  Generating adversarial examples for natural language is fundamentally different from generating adversarial examples for images.  ... 
doi:10.18653/v1/2020.coling-main.585 fatcat:7wpnzeeywbbaldc5qq45ank4ym

Contrasting Human- and Machine-Generated Word-Level Adversarial Examples for Text Classification

Maximilian Mozes, Max Bartolo, Pontus Stenetorp, Bennett Kleinberg, Lewis Griffin
2021 Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing   unpublished
Our findings suggest that human-generated adversarial examples are not more able than the best algorithms to generate natural-reading, sentimentpreserving examples, though they do so by being much more  ...  We analyze how human-generated adversarial examples compare to the recently proposed TEXTFOOLER, GENETIC, BAE and SEMEMEPSO attack algorithms on the dimensions naturalness, preservation of sentiment, grammaticality  ...  These findings show that under similar constraints, machine-generated, word-level adversarial examples are comparable to human-generated ones with respect to their naturalness and grammaticality.  ... 
doi:10.18653/v1/2021.emnlp-main.651 fatcat:w6x4ddxujbczjoiaxcdwmc2tlu

Feature-Based Adversarial Training for Deep Learning Models Resistant to Transferable Adversarial Examples

Gwonsang RYU, Daeseon CHOI
2022 IEICE transactions on information and systems  
Adversaries may generate adversarial examples using the property of transferability, in which adversarial examples that deceive one model can also deceive other models because adversaries do not obtain  ...  We propose a novel adversarial training method to train DNNs to be robust against transferable adversarial examples and maximize their classification accuracy for natural images.  ...  For example, we generate adversarial examples to deceive the model that is naturally trained by Xception and then input the generated adversarial examples into the naturally trained model as well as those  ... 
doi:10.1587/transinf.2021edp7198 fatcat:gu2mrqe54vb5nctck74ivaazpq

Push Stricter to Decide Better: A Class-Conditional Feature Adaptive Framework for Improving Adversarial Robustness [article]

Jia-Li Yin, Lehui Xie, Wanqing Zhu, Ximeng Liu, Bo-Hao Chen
2021 arXiv   pre-print
adversarial examples, leading to a dramatic decrease in natural accuracy.  ...  class-conditional feature adaption across natural data and adversarial examples.  ...  (a) Shift between natural and adversarial data. Note that the adversarial examples are generated by adding imperceptible perturbations over the natural data.  ... 
arXiv:2112.00323v1 fatcat:qf4vdsli6jbadhruy2kan7nfau

ReabsNet: Detecting and Revising Adversarial Examples [article]

Jiefeng Chen, Zihang Meng, Changtian Sun, Wei Tang, Yinglun Zhu
2017 arXiv   pre-print
Critically, instead of simply rejecting adversarial examples, we revise them to get their true labels.  ...  The approach is to augment an existing classification network with a guardian network to detect if a sample is natural or has been adversarially perturbed.  ...  Subsequently, we generate adversarial data for each natural examples in the training dataset through one of the attacking methods discussed in Section 3.2.  ... 
arXiv:1712.08250v1 fatcat:ozxpdvlkbfgkncowqot7wvtw6a
« Previous Showing results 1 — 15 out of 172,110 results