English(EN) Private LLM copies can agree on a secret key over a public channel

私有LLM副本通过公共信道就密钥达成一致

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-09 07:47

研究人员已证明，大型语言模型（LLM）的独立私有副本仅通过公共通信信道即可协作建立共享密钥。该方法利用了隐写术，即将信息隐藏在看似无害的数据传输中。即使LLM实例无法直接通信，该技术也能实现安全的密钥协商，可能为私有通信协议开辟新途径。 AI

影响展示了一种隔离LLM实例之间安全通信的新颖方法，可能影响私有AI交互。

排序理由学术论文，详细介绍了一种新颖的LLM通信方法。[lever_c_demoted from research: ic=1 ai=1.0]

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: 39 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 学术论文，详细介绍了一种新颖的LLM通信方法。[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: safety, other; 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。

LessWrong (AI tag) TIER_1 English(EN) · matthew-ritch · 2026-11-09 07:47

私有大模型副本可通过公开频道就密钥达成一致

Cross-posted from <a href="https://matthewritch.com/blog/2026/11/08/LLM-Preferences-Private-Keys-Fingerprinting/">my blog</a>. Epistemic status exploratory, no running protocol, theoretical estimates from a single experiment and 4 closed-weights models…