English(EN) Known By Their Actions: Fingerprinting LLM Browser Agents via UI Traces

研究人员通过用户界面交互痕迹指纹识别大语言模型浏览器代理

作者 PulseAugur 编辑部 · [1 个来源] · 2026-05-14 12:55

研究人员开发了一种方法，通过分析大型语言模型（LLM）驱动的网络浏览代理的交互模式和时序来识别其具体身份。该技术已在14种不同的LLM和各种网络环境中得到验证，能够以高达96%的准确率识别底层模型。研究发现，即使在操作之间随机延迟的情况下，仍能推断出代理的身份，这表明存在基于已知模型漏洞进行定向攻击的潜在安全风险。研究人员已发布了他们的追踪工具和代理痕迹数据集，以促进进一步研究。 AI

影响这项研究突显了大语言模型代理潜在的安全漏洞，使得通过识别其网络浏览行为来针对特定模型进行攻击成为可能。

排序理由详细介绍一种识别大语言模型代理新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 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: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 详细介绍一种识别大语言模型代理新方法的学术论文。[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: 180 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.AI TIER_1 English(EN) · Chris Russell · 2026-05-14 12:55

通过UI痕迹识别LLM浏览器代理：观其行而知其名

As LLM-based agents increasingly browse the web on users' behalf, a natural question arises: can websites passively identify which underlying model powers an agent? Doing so would represent a significant security risk, enabling targeted attacks tailored to known model vulnerabili…

本文如何被排名

报道来源 [1]

通过UI痕迹识别LLM浏览器代理：观其行而知其名

相关实体

相关话题