{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/generating-natural-adversarial-examples","title":"Generating Natural Adversarial Examples","arxiv_id":"1710.11342","date":"2017-10-31","proceeding":"ICLR 2018 1","authors":["Zhengli Zhao","Dheeru Dua","Sameer Singh"],"abstract":"Due to their complex nature, it is hard to characterize the ways in which\nmachine learning models can misbehave or be exploited when deployed. Recent\nwork on adversarial examples, i.e. inputs with minor perturbations that result\nin substantially different model predictions, is helpful in evaluating the\nrobustness of these models by exposing the adversarial scenarios where they\nfail. However, these malicious perturbations are often unnatural, not\nsemantically meaningful, and not applicable to complicated domains such as\nlanguage. In this paper, we propose a framework to generate natural and legible\nadversarial examples that lie on the data manifold, by searching in semantic\nspace of dense and continuous data representation, utilizing the recent\nadvances in generative adversarial networks. We present generated adversaries\nto demonstrate the potential of the proposed approach for black-box classifiers\nfor a wide range of applications such as image classification, textual\nentailment, and machine translation. We include experiments to show that the\ngenerated adversaries are natural, legible to humans, and useful in evaluating\nand analyzing black-box classifiers.","url_abs":"http://arxiv.org/abs/1710.11342v2","url_pdf":"http://arxiv.org/pdf/1710.11342v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"generating-natural-adversarial-examples","repo_url":"https://github.com/zhengliz/natural-adversary","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"adversarial-attack","task_name":"Adversarial Attack"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"natural-language-inference","task_name":"Natural Language Inference"},{"task_slug":"translation","task_name":"Translation"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1710.11342","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}