cissp-llmbench
收藏资源简介:
CISSP-LLMBench 是一个专注于网络安全领域的大型语言模型安全评估基准数据集。其核心内容包含与网络安全相关的红队提示,旨在测试和提升语言模型在应对涉及安全威胁、漏洞或恶意行为查询时的鲁棒性与安全性。数据集规模在1万到10万个样本之间,语言为英语。它设计用于防御性AI安全研究,适用于文本生成和文本分类等任务场景,特别是评估模型对齐能力、构建AI安全基准以及进行负责任的网络安全能力红队测试。数据集名称中的CISSP暗示其内容可能与网络安全知识体系(如CISSP认证相关领域)相关联。使用本数据集需同意仅将其用于防御性研究目的,不得用于恶意用途。
CISSP-LLMBench is a large language model security evaluation benchmark dataset focused on the cybersecurity domain. Its core content consists of red-teaming prompts related to cybersecurity, designed to test and enhance the robustness and safety of LLMs when responding to queries involving security threats, vulnerabilities, or malicious behaviors. The dataset scale ranges from 10,000 to 100,000 samples, and the language is English. It is designed for defensive AI security research, suitable for tasks such as text generation and text classification, particularly for evaluating model alignment capabilities, building AI security benchmarks, and conducting responsible red-team testing of cybersecurity capabilities. The CISSP in the dataset name suggests a potential connection to cybersecurity knowledge systems, such as those related to CISSP certification. Use of this dataset requires agreement to employ it solely for defensive research purposes and not for any malicious applications.
数据集概述:CISSP-LLMBench
- 名称:CISSP-LLMBench
- 许可证:Creative Commons Attribution 4.0 International (CC-BY-4.0)
- 语言:英语 (en)
- 任务类别:文本生成、文本分类
- 标签:网络安全、大语言模型安全、红队测试、CISSP(认证信息系统安全专业人士)、对齐、AI安全基准
- 规模:样本数量区间为 10,000 至 100,000 (10K < n < 100K)
内容与用途
- 内容性质:包含与网络安全相关的红队测试提示(red-teaming prompts)。
- 研究目的:专为防御性AI安全研究设计,用户需承诺不将这些提示用于恶意目的。
- 访问限制:需同意负责任使用条款,并填写所属机构/隶属关系方可获取。




