{"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/nag-network-for-adversary-generation","title":"NAG: Network for Adversary Generation","arxiv_id":"1712.03390","date":"2017-12-09","proceeding":"CVPR 2018 6","authors":["Konda Reddy Mopuri","Utkarsh Ojha","Utsav Garg","R. Venkatesh Babu"],"abstract":"Adversarial perturbations can pose a serious threat for deploying machine\nlearning systems. Recent works have shown existence of image-agnostic\nperturbations that can fool classifiers over most natural images. Existing\nmethods present optimization approaches that solve for a fooling objective with\nan imperceptibility constraint to craft the perturbations. However, for a given\nclassifier, they generate one perturbation at a time, which is a single\ninstance from the manifold of adversarial perturbations. Also, in order to\nbuild robust models, it is essential to explore the manifold of adversarial\nperturbations. In this paper, we propose for the first time, a generative\napproach to model the distribution of adversarial perturbations. The\narchitecture of the proposed model is inspired from that of GANs and is trained\nusing fooling and diversity objectives. Our trained generator network attempts\nto capture the distribution of adversarial perturbations for a given classifier\nand readily generates a wide variety of such perturbations. Our experimental\nevaluation demonstrates that perturbations crafted by our model (i) achieve\nstate-of-the-art fooling rates, (ii) exhibit wide variety and (iii) deliver\nexcellent cross model generalizability. Our work can be deemed as an important\nstep in the process of inferring about the complex manifolds of adversarial\nperturbations.","url_abs":"http://arxiv.org/abs/1712.03390v2","url_pdf":"http://arxiv.org/pdf/1712.03390v2.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":"nag-network-for-adversary-generation","repo_url":"https://github.com/val-iisc/nag","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1712.03390","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}