大语言模型能提供安全与隐私建议吗？衡量LLMs驳斥错误观念的能力
Can Large Language Models Provide Security & Privacy Advice? Measuring the Ability of LLMs to Refute Misconceptions
AI总结：
该研究构建了含百余条安全隐私错误观念的数据集，通过多策略测试Bard与ChatGPT，发现二者驳斥错误观念的错误率较高，且常提供无效或无关来源URL，揭示了LLMs提供安全隐私建议的可靠性不足问题。
摘要：
用户会从包括可信网站和内容分享平台在内的在线资源中寻求安全与隐私（S&P）建议，这些资源能帮助用户理解安全与隐私相关的技术和工具，并提出可执行的策略。大语言模型（LLMs）近来已成为可信的信息来源，但其准确性和正确性却受到质疑。已有研究指出了LLMs在回答选择题方面的不足，以及用户可能在无意中绕过模型限制（例如生成有害内容）的问题，但目前尚未充分探究LLMs提供可靠安全与隐私建议的能力。本文中，我们衡量了LLMs驳斥普通大众所持的常见安全与隐私错误观念的能力。我们首先研究了近期的学术文献，整理出一个包含六大主题、超过一百条安全与隐私相关错误观念的数据集。随后，我们向两款主流LLMs（Bard和ChatGPT）发起查询，并制定了标注指南来评估它们对这些错误观念的回应。为全面评估其回应，我们进一步采用了三种策略：对每条错误观念多次查询、生成并查询其改写版本、索要回应的来源URL。两款模型平均存在21.3%的不可忽视的错误率，会错误地支持常见的安全与隐私错误观念。当我们用相同或改写后的错误观念重复查询LLMs时，错误率上升至32.6%。我们还发现，模型可能会部分支持某一错误观念，或是态度含糊，拒绝对错误观念表明明确立场。我们对回应信息来源的探究显示，LLMs容易提供无效URL（Bard为21.2%，ChatGPT为67.7%）或是指向无关来源（Bard返回的占比为44.2%，ChatGPT为18.3%）。
英文摘要：
Users seek security & privacy (S&P) advice from online resources, including trusted websites and content-sharing platforms. These resources help users understand S&P technologies and tools and suggest actionable strategies. Large Language Models (LLMs) have recently emerged as trusted information sources. However, their accuracy and correctness have been called into question. Prior research has outlined the shortcomings of LLMs in answering multiple-choice questions and user ability to inadvertently circumvent model restrictions (e.g., to produce toxic content). Yet, the ability of LLMs to provide reliable S&P advice is not well-explored. In this paper, we measure their ability to refute popular S&P misconceptions that the general public holds. We first study recent academic literature to curate a dataset of over a hundred S&P-related misconceptions across six different topics. We then query two popular LLMs (Bard and ChatGPT) and develop a labeling guide to evaluate their responses to these misconceptions. To comprehensively evaluate their responses, we further apply three strategies: query each misconception multiple times, generate and query their paraphrases, and solicit source URLs of the responses. Both models demonstrate, on average, a 21.3% non-negligible error rate, incorrectly supporting popular S&P misconceptions. The error rate increases to 32.6% when we repeatedly query LLMs with the same or paraphrased misconceptions. We also expose that models may partially support a misconception or remain noncommittal, refusing a firm stance on misconceptions. Our exploration of information sources for responses revealed that LLMs are susceptible to providing invalid URLs (21.2% for Bard and 67.7% for ChatGPT) or point to unrelated sources (44.2% returned by Bard and 18.3% by ChatGPT).
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。
