shaswatamitra/falcon-snort-cti-rule
收藏资源简介:
FALCON SNORT CTI ↔ 真实规则数据集是一个专门用于网络安全领域的数据集,它包含网络威胁情报(CTI)描述与其对应的真实SNORT入侵检测系统(IDS)规则,以及由大型语言模型(LLM)生成的诱饵规则(硬负样本)。该数据集旨在作为LLM规则生成评估和检索测试的基准测试集,用于评估LLM根据CTI描述生成SNORT规则的能力,以及测试对比性微调的句子编码器在检索任务中的性能。数据集包含802个CTI描述,总共3,063个诱饵规则,平均每个CTI有3.82个诱饵规则,范围在2到6个之间。数据集为单一测试分割,是保留的测试集,不可用于训练。
The FALCON SNORT CTI ↔ Ground-Truth Rule Dataset is a cybersecurity dataset that pairs cyber-threat-intelligence (CTI) descriptions with their ground-truth SNORT IDS rules and LLM-generated decoy rules (hard-negative look-alikes). It serves as a held-out test benchmark for evaluating LLM rule-generation and retrieval testing of contrastively fine-tuned sentence encoders. The dataset includes 802 CTIs, 3,063 total decoys (mean 3.82 per CTI, range 2-6), and has a single split: test, which should not be used for training.




