CyberSecEval 2:一个面向大型语言模型的广泛网络安全评估套件
CyberSecEval 2: A Wide-Ranging Cybersecurity Evaluation Suite for Large Language Models
AI总结:
本文提出CyberSecEval 2,一个用于量化LLM安全风险与能力的基准,新增提示词注入与代码解释器滥用测试,并引入误拒率(FRR)以衡量安全与实用性权衡,发现现有模型在防范攻击和漏洞利用生成方面仍有不足。
摘要:
大型语言模型(LLMs)引入了新的安全风险,但鲜有全面的评估套件来衡量和降低这些风险。我们提出了BenchmarkName,一个用于量化LLM安全风险与能力的新颖基准。我们引入了两个新的测试领域:提示词注入和代码解释器滥用。我们评估了多个最先进的(SOTA)LLM,包括GPT-4、Mistral、Meta Llama 3 70B-Instruct和Code Llama。我们的结果表明,通过条件设定消除攻击风险仍是一个未解决的问题;例如,所有测试模型在26%到41%的提示词注入测试中取得了成功。我们进一步引入了安全-实用性权衡:通过条件设定使LLM拒绝不安全的提示词,可能导致LLM错误地拒绝回答良性提示词,从而降低实用性。我们提议使用误拒率(False Refusal Rate, FRR)来量化这一权衡。作为示例,我们引入了一个新颖的测试集来量化网络攻击帮助性风险的FRR。我们发现许多LLM能够成功遵从“边界”良性请求,同时仍拒绝大多数不安全请求。最后,我们量化了LLM在自动化一项核心网络安全任务(即利用软件漏洞)方面的实用性。这很重要,因为LLM的攻击能力备受关注;我们通过为四个代表性问题创建新颖的测试集来量化这一点。我们发现具备编码能力的模型表现优于不具备该能力的模型,但LLM要熟练生成漏洞利用程序仍需进一步研究。我们的代码是开源的,可用于评估其他LLM。
英文摘要:
Large language models (LLMs) introduce new security risks, but there are few comprehensive evaluation suites to measure and reduce these risks. We present BenchmarkName, a novel benchmark to quantify LLM security risks and capabilities. We introduce two new areas for testing: prompt injection and code interpreter abuse. We evaluated multiple state-of-the-art (SOTA) LLMs, including GPT-4, Mistral, Meta Llama 3 70B-Instruct, and Code Llama. Our results show that conditioning away risk of attack remains an unsolved problem; for example, all tested models showed between 26% and 41% successful prompt injection tests. We further introduce the safety-utility tradeoff: conditioning an LLM to reject unsafe prompts can cause the LLM to falsely reject answering benign prompts, which lowers utility. We propose quantifying this tradeoff using False Refusal Rate (FRR). As an illustration, we introduce a novel test set to quantify FRR for cyberattack helpfulness risk. We find many LLMs able to successfully comply with "borderline" benign requests while still rejecting most unsafe requests. Finally, we quantify the utility of LLMs for automating a core cybersecurity task, that of exploiting software vulnerabilities. This is important because the offensive capabilities of LLMs are of intense interest; we quantify this by creating novel test sets for four representative problems. We find that models with coding capabilities perform better than those without, but that further work is needed for LLMs to become proficient at exploit generation. Our code is open source and can be used to evaluate other LLMs.
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。