研究人员开发了一种名为MIST的新方法，用于检测在微调过程中嵌入深度神经网络（DNN）中的恶意木马。MIST分析模型内部表示的谱变化，以识别表明木马攻击的偏差。这种方法将木马检测视为一个回归问题，并且与现有方法相比，即使在事先不知道攻击细节的情况下，也表现出更高的准确性。 AI
影响 引入了一种新颖的技术，以增强AI模型在开发过程中抵御复杂攻击的安全性。
排序理由 学术论文，详细介绍了一种检测AI模型安全漏洞的新方法。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2605.21146v1 Announce Type: cross Abstract: Modern DNNs are repeatedly fine-tuned to incorporate new data and functionality. This evolutionary workflow introduces a security risk when updated data cannot be fully trusted, as adversaries may implant Trojans during fine-tunin…
Modern DNNs are repeatedly fine-tuned to incorporate new data and functionality. This evolutionary workflow introduces a security risk when updated data cannot be fully trusted, as adversaries may implant Trojans during fine-tuning. We present MIST, a Trojan detection approach th…