arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.09452cs.CRcs.CL
收藏

CyberThreat-Eval: 大型语言模型能否自动化现实中的威胁研究?

CyberThreat-Eval: Can Large Language Models Automate Real-World Threat Research?

Microsoft Research(微软研究院);Microsoft(微软)

Xiangsen Chen, Xuan Feng, Shuo Chen, Matthieu Maitre, Sudipto Rakshit, Diana Duvieilh, Ashley Picone, Nan Tang

AI总结:

CyberThreat-Eval通过专家标注的基准测试评估大型语言模型在威胁情报工作流程中的实际表现,揭示其在复杂细节处理和信息区分上的不足。

摘要:

分析大量数据中的开源情报(OSINT)对于撰写和发布全面的威胁情报(CTI)报告至关重要。这一过程通常遵循一个三阶段的工作流程——筛选、深入搜索和情报报告起草。尽管大型语言模型(LLMs)提供了一条实现自动化的有希望的途径,但现有的基准测试仍然存在局限性。这些基准测试通常由不反映现实世界分析师工作流程的任务组成。例如,人类分析师很少会以多项选择题的形式接收任务。此外,现有的基准测试往往依赖于以模型为中心的度量标准,强调词汇重叠而不是安全分析师所需的可操作、详细的见解。而且,它们通常无法覆盖完整的三阶段工作流程。为了解决这些问题,我们引入了CyberThreat-Eval,该基准测试从一家世界领先的公司日常CTI工作流程中收集而来。这个由专家标注的基准测试评估LLMs在上述三个阶段的实践中任务。它利用以分析师为中心的度量标准,测量事实准确性、内容质量和运营成本。使用此基准测试的评估揭示了当前LLMs的局限性。例如,LLMs常常缺乏处理复杂细节所需的细微专业知识,并且在区分正确和错误信息方面存在困难。为了解决这些挑战,CTI工作流程结合了外部真实数据库和人类专家知识。TRA允许人类专家通过迭代提供反馈以实现持续改进。代码可在https://github.com/xschen-beb/CyberThreat-Eval和https://huggingface.co/datasets/xse/CyberThreat-Eval上获得。

英文摘要:

Analyzing Open Source Intelligence (OSINT) from large volumes of data is critical for drafting and publishing comprehensive CTI reports. This process usually follows a three-stage workflow -- triage, deep search and TI drafting. While Large Language Models (LLMs) offer a promising route toward automation, existing benchmarks still have limitations. These benchmarks often consist of tasks that do not reflect real-world analyst workflows. For example, human analysts rarely receive tasks in the form of multiple-choice questions. Also, existing benchmarks often rely on model-centric metrics that emphasize lexical overlap rather than actionable, detailed insights essential for security analysts. Moreover, they typically fail to cover the complete three-stage workflow. To address these issues, we introduce CyberThreat-Eval, which is collected from the daily CTI workflow of a world-leading company. This expert-annotated benchmark assesses LLMs on practical tasks across all three stages as mentioned above. It utilizes analyst-centric metrics that measure factual accuracy, content quality, and operational costs. Our evaluation using this benchmark reveals important insights into the limitations of current LLMs. For example, LLMs often lack the nuanced expertise required to handle complex details and struggle to distinguish between correct and incorrect information. To address these challenges, the CTI workflow incorporates both external ground-truth databases and human expert knowledge. TRA allows human experts to iteratively provide feedback for continuous improvement. The code is available at \href{https://github.com/xschen-beb/CyberThreat-Eval}{\texttt{GitHub}} and \href{https://huggingface.co/datasets/xse/CyberThreat-Eval}{\texttt{HuggingFace}}.

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。

↑