English(EN) Black-Box Membership Inference via Word-Level Probability Estimation

新的黑盒MIA技术评估LLM训练数据隐私

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-11 04:00

研究人员开发了一种名为词级概率MIA（WPMIA）的新方法，即使只能访问文本续写，也能检测文本是否属于大型语言模型的训练数据。该技术通过蒙特卡洛采样和核平滑来估计词级生成概率，然后将这些概率聚合为序列级似然估计器。WPMIA在GPT-5 Chat、Gemini 2.5-Flash和Claude 4.5 Haiku等专有模型上进行了测试，在5%的假阳性率（FPR）下实现了42.0%的真阳性率（TPR），证明了其在严格的黑盒隐私审计中的有效性。 AI

影响这项新的隐私审计技术可能会迫使LLM开发者加强专有模型的数据保护措施。

排序理由该集群包含一篇详细介绍LLM隐私审计新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv stat.ML 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: 27 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍LLM隐私审计新方法的学术论文。[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: 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 stat.ML TIER_1 English(EN) · Shengjie Niu, Yeheng Ge, Jian Huang · 2026-09-11 04:00

通过词级概率估计实现的黑盒成员推理

arXiv:2609.10611v1 Announce Type: cross Abstract: Membership inference attacks (MIAs) have emerged as critical tools for auditing privacy risks in large language models (LLMs), aiming to determine whether a given text was included in a model's training corpus. However, most exist…

本文如何被排名

报道来源 [1]

通过词级概率估计实现的黑盒成员推理

相关实体

相关话题