{"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/black-box-adversarial-attacks-with-limited","title":"Black-box Adversarial Attacks with Limited Queries and Information","arxiv_id":"1804.08598","date":"2018-04-23","proceeding":"ICML 2018 7","authors":["Andrew Ilyas","Logan Engstrom","Anish Athalye","Jessy Lin"],"abstract":"Current neural network-based classifiers are susceptible to adversarial\nexamples even in the black-box setting, where the attacker only has query\naccess to the model. In practice, the threat model for real-world systems is\noften more restrictive than the typical black-box model where the adversary can\nobserve the full output of the network on arbitrarily many chosen inputs. We\ndefine three realistic threat models that more accurately characterize many\nreal-world classifiers: the query-limited setting, the partial-information\nsetting, and the label-only setting. We develop new attacks that fool\nclassifiers under these more restrictive threat models, where previous methods\nwould be impractical or ineffective. We demonstrate that our methods are\neffective against an ImageNet classifier under our proposed threat models. We\nalso demonstrate a targeted black-box attack against a commercial classifier,\novercoming the challenges of limited query access, partial information, and\nother practical issues to break the Google Cloud Vision API.","url_abs":"http://arxiv.org/abs/1804.08598v3","url_pdf":"http://arxiv.org/pdf/1804.08598v3.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":"black-box-adversarial-attacks-with-limited","repo_url":"https://github.com/labsix/limited-blackbox-attacks","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"black-box-adversarial-attacks-with-limited","repo_url":"https://github.com/mllab-adv-attack/lazy-attack","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1804.08598","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}