研究人员开发了一种名为威胁感知频率解耦（TaFD）的新型防御框架，以提高深度学习模型的对抗鲁棒性。TaFD通过将联合对抗训练重新构建为频域方法来应对异构攻击（如$\ell_p$-有界攻击和语义攻击）的挑战。该框架通过无监督聚类识别威胁域，然后使用频率条件卷积将样本路由到专门的专家，从而缓解优化冲突并增强均衡鲁棒性。 AI
影响 增强模型对各种对抗性攻击的抵御能力，可能提高AI系统在安全敏感应用中的可靠性。
排序理由 该集群包含一篇研究论文，详细介绍了用于提高深度学习模型对抗鲁棒性的新技术框架。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2606.17540v1 Announce Type: new Abstract: Multi-threat robustness remains a fundamental challenge in deep learning. Although joint adversarial training (JAT) is widely adopted, it suffers from negative transfer under heterogeneous threats, particularly between $\ell_p$-boun…
Multi-threat robustness remains a fundamental challenge in deep learning. Although joint adversarial training (JAT) is widely adopted, it suffers from negative transfer under heterogeneous threats, particularly between $\ell_p$-bounded and semantic attacks. Through first-order gr…