{"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/fine-pruning-defending-against-backdooring","title":"Fine-Pruning: Defending Against Backdooring Attacks on Deep Neural Networks","arxiv_id":"1805.12185","date":"2018-05-30","proceeding":null,"authors":["Kang Liu","Brendan Dolan-Gavitt","Siddharth Garg"],"abstract":"Deep neural networks (DNNs) provide excellent performance across a wide range\nof classification tasks, but their training requires high computational\nresources and is often outsourced to third parties. Recent work has shown that\noutsourced training introduces the risk that a malicious trainer will return a\nbackdoored DNN that behaves normally on most inputs but causes targeted\nmisclassifications or degrades the accuracy of the network when a trigger known\nonly to the attacker is present. In this paper, we provide the first effective\ndefenses against backdoor attacks on DNNs. We implement three backdoor attacks\nfrom prior work and use them to investigate two promising defenses, pruning and\nfine-tuning. We show that neither, by itself, is sufficient to defend against\nsophisticated attackers. We then evaluate fine-pruning, a combination of\npruning and fine-tuning, and show that it successfully weakens or even\neliminates the backdoors, i.e., in some cases reducing the attack success rate\nto 0% with only a 0.4% drop in accuracy for clean (non-triggering) inputs. Our\nwork provides the first step toward defenses against backdoor attacks in deep\nneural networks.","url_abs":"http://arxiv.org/abs/1805.12185v1","url_pdf":"http://arxiv.org/pdf/1805.12185v1.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":"fine-pruning-defending-against-backdooring","repo_url":"https://github.com/Alaqian/Pruning-Defense-Against-Backdooring-Attacks-on-Deep-Neural-Networks","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"fine-pruning-defending-against-backdooring","repo_url":"https://github.com/nicholasbennet/ml-cybersec-proj","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"fine-pruning-defending-against-backdooring","repo_url":"https://github.com/unispac/subnet-replacement-attack","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[],"methods":[{"method_slug":"pruning","method_name":"Pruning"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1805.12185","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}