研究人员开发了MemPrivacy，一个旨在保护利用边缘计算和云计算的AI代理中的敏感用户数据的新框架。该系统采用本地可逆假名化，在将私人信息发送到云端之前，用类型化的占位符替换它。云模型随后可以语义完整地处理数据，并在收到响应后在用户的设备上恢复原始信息。这种方法旨在在不损害用户隐私的情况下，保持个性化AI内存的实用性。 AI
影响 通过实现个性化内存的安全使用而不暴露敏感用户数据，增强了AI代理的隐私性。
排序理由 该集群描述了一个新的研究框架及其解决问题的方法。
AI 生成摘要 · Google Gemini · 来自 4 个来源。 我们如何撰写摘要 →
<p>As LLM-powered agents move from research to production, one design tension is becoming harder to ignore: the more useful cloud-hosted memory becomes, the more private user data it exposes. Researchers from MemTensor (Shanghai), HONOR Device and Tongji University have introduce…
MemPrivacy is a new edge-cloud framework that protects user data in AI memory systems by replacing sensitive values with typed placeholders before cloud transmission. Developed by MemTensor, HONOR and Tongji University, it achieves 85.97% F1 on privacy extraction benchmarks, outp…
📰 MemPrivacy Framework (2026): AI Data Protection via Reversible Pseudonymization A new framework called MemPrivacy uses local reversible pseudonymization to protect sensitive user data in edge-cloud AI agents without sacrificing memory utility. Developed by researchers from MemT…
📰 MemPrivacy: 2026'de Yapay Zeka Gizliliğini Edge-Cloud ile Koruyan Devrim MemTensor ekibi, yapay zeka asistanlarının uzun vadeli kişiselleştirme yeteneklerini korurken kullanıcı verilerini koruyan devrimsel bir çerçeve geliştirdi. MemPrivacy adı verilen sistem, hassas verileri y…