English(EN) When Routing Reveals Membership: Privacy Leakage from MoE Router Telemetry

新攻击利用MoE路由器遥测数据进行隐私泄露

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-09 04:00

研究人员开发了一种新的成员推断攻击，该攻击利用了混合专家（MoE）语言模型的遥测数据。这种名为“路由器增强成员推断攻击”的方法，结合了传统的输出信号和聚合的路由特征，以确定特定的数据样本是否在模型的微调过程中被使用。研究结果表明，即使在实施了隐私保护措施的情况下，这种遥测数据也能在各种MoE架构和数据域中持续提高成员推断攻击的准确性。 AI

影响突显了MoE模型中新的隐私漏洞，可能影响遥测数据的处理和安全方式。

排序理由详细介绍新隐私攻击方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 18 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 详细介绍新隐私攻击方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, safety; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Yixin Tan, Jiayang Liu, Lu Sun, Yuke Hu, Zheng Li, Rui Wen · 2026-10-09 04:00

路由泄露成员身份：MoE路由器遥测数据中的隐私泄露

arXiv:2610.10616v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) language models produce routing information during inference that may be logged or exposed for monitoring, debugging, load analysis, and safety auditing. Unlike ordinary model outputs, this telemetry reveals…

本文如何被排名

报道来源 [1]

路由泄露成员身份：MoE路由器遥测数据中的隐私泄露

相关实体

相关话题