{"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/improved-network-robustness-with-adversary","title":"Improved Network Robustness with Adversary Critic","arxiv_id":"1810.12576","date":"2018-10-30","proceeding":"NeurIPS 2018 12","authors":["Alexander Matyasko","Lap-Pui Chau"],"abstract":"Ideally, what confuses neural network should be confusing to humans. However,\nrecent experiments have shown that small, imperceptible perturbations can\nchange the network prediction. To address this gap in perception, we propose a\nnovel approach for learning robust classifier. Our main idea is: adversarial\nexamples for the robust classifier should be indistinguishable from the regular\ndata of the adversarial target. We formulate a problem of learning robust\nclassifier in the framework of Generative Adversarial Networks (GAN), where the\nadversarial attack on classifier acts as a generator, and the critic network\nlearns to distinguish between regular and adversarial images. The classifier\ncost is augmented with the objective that its adversarial examples should\nconfuse the adversary critic. To improve the stability of the adversarial\nmapping, we introduce adversarial cycle-consistency constraint which ensures\nthat the adversarial mapping of the adversarial examples is close to the\noriginal. In the experiments, we show the effectiveness of our defense. Our\nmethod surpasses in terms of robustness networks trained with adversarial\ntraining. Additionally, we verify in the experiments with human annotators on\nMTurk that adversarial examples are indeed visually confusing. Codes for the\nproject are available at https://github.com/aam-at/adversary_critic.","url_abs":"http://arxiv.org/abs/1810.12576v1","url_pdf":"http://arxiv.org/pdf/1810.12576v1.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":"improved-network-robustness-with-adversary","repo_url":"https://github.com/aam-at/adversary_critic","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"adversarial-attack","task_name":"Adversarial Attack"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1810.12576","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}