{"slug": "two-sides-of-the-same-coin-learning-the-backdoor-to-remove-the-backdoor", "title": "Two Sides of the Same Coin: Learning the Backdoor to Remove the Backdoor", "summary": "Researchers at arXiv introduced HARVEY, a defense against neural backdoor attacks that learns a reference model for poisonous samples instead of benign ones. By identifying poisonous samples more accurately than prior methods, HARVEY achieves near-perfect backdoor removal with minimal loss in natural accuracy, outperforming existing defenses across various attacks, datasets, and architectures.", "body_md": "arXiv:2607.05748v1 Announce Type: new\nAbstract: The community has recently developed various training-time defenses to counter neural backdoors introduced through data poisoning. In light of the observation that a model learns poisonous samples responsible for the backdoor easier than benign samples, these approaches either use a fixed threshold of the training loss for splitting or iteratively learn a reference model as an oracle for identifying benign samples. In particular, the latter has proven effective for anti-backdoor learning.\nOur method, HARVEY, leverages a similar yet crucially different technique: learning an oracle for poisonous rather than benign samples. Learning a backdoored reference model is significantly easier than learning a reference model on benign data. Consequently, we can identify poisonous samples much more accurately than related work identifies benign samples. This crucial difference enables near-perfect backdoor removal as we demonstrate in our evaluation. HARVEY substantially outperforms related approaches across attack types, datasets, and architectures, lowering the attack success rate to the very minimum at a negligible loss in natural accuracy. The figure below shows an overview of our methods working principle.", "url": "https://wpnews.pro/news/two-sides-of-the-same-coin-learning-the-backdoor-to-remove-the-backdoor", "canonical_source": "https://www.machinebrief.com/news/two-sides-of-the-same-coin-learning-the-backdoor-to-remove-t-v7d4", "published_at": "2026-07-08 04:00:00+00:00", "updated_at": "2026-07-08 04:18:55.656737+00:00", "lang": "en", "topics": ["ai-safety", "neural-networks", "machine-learning"], "entities": ["HARVEY", "arXiv"], "alternates": {"html": "https://wpnews.pro/news/two-sides-of-the-same-coin-learning-the-backdoor-to-remove-the-backdoor", "markdown": "https://wpnews.pro/news/two-sides-of-the-same-coin-learning-the-backdoor-to-remove-the-backdoor.md", "text": "https://wpnews.pro/news/two-sides-of-the-same-coin-learning-the-backdoor-to-remove-the-backdoor.txt", "jsonld": "https://wpnews.pro/news/two-sides-of-the-same-coin-learning-the-backdoor-to-remove-the-backdoor.jsonld"}}