模态迷失：评估基于文本的成员推断攻击在大型多模态模型中的有效性
Lost in Modality: Evaluating the Effectiveness of Text-Based Membership Inference Attacks on Large Multimodal Models
AI总结：
本文评估了基于文本的成员推断攻击（MIAs）在多模态模型中的有效性，发现其在分布内设置中表现相似，而在分布外设置中视觉输入起到正则化作用，有效掩盖了成员信号。
摘要：
大型多模态语言模型（MLLMs）正在成为越来越多应用中的基础工具。因此，理解这些系统中的训练数据泄漏变得越来越关键。基于对数概率的成员推断攻击（MIAs）已成为评估大型语言模型（LLMs）数据暴露的常用方法，但其在MLLMs中的效果仍不明确。我们首次全面评估了将这些基于文本的MIA方法扩展到多模态设置的效果。在DeepSeek-VL和InternVL模型家族中，我们进行了视觉与文本（V+T）和纯文本（T-only）条件下的实验，结果显示，在分布内设置中，基于logit的MIAs在不同配置中表现相当，略有V+T优势。相反，在分布外设置中，视觉输入起到正则化作用，有效掩盖了成员信号。
英文摘要：
Large Multimodal Language Models (MLLMs) are emerging as one of the foundational tools in an expanding range of applications. Consequently, understanding training-data leakage in these systems is increasingly critical. Log-probability-based membership inference attacks (MIAs) have become a widely adopted approach for assessing data exposure in large language models (LLMs), yet their effect in MLLMs remains unclear. We present the first comprehensive evaluation of extending these text-based MIA methods to multimodal settings. Our experiments under vision-and-text (V+T) and text-only (T-only) conditions across the DeepSeek-VL and InternVL model families show that in in-distribution settings, logit-based MIAs perform comparably across configurations, with a slight V+T advantage. Conversely, in out-of-distribution settings, visual inputs act as regularizers, effectively masking membership signals.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
