knowurknottty/Cybersecurity-Dataset-Fenrir-v2.1

网络安全防御指令调优数据集（v2.1）是一个包含99,870条高质量system/user/assistant三元组的英语数据集，专为防御性、对齐安全的网络安全监督微调（SFT）训练设计。数据集覆盖了多个网络安全标准和框架，包括OWASP Top 10、MITRE ATT&CK、NIST CSF、CIS Controls等，并特别关注云安全、DevSecOps、现代身份验证（OAuth 2/OIDC/SAML）、SSL/TLS、密码学和AI安全。数据集通过Apache-2.0许可发布，适合商业用途。数据集的创建过程包括源数据收集、提取、主题过滤、指令合成和多层质量检查，确保数据的高质量和安全性。数据集还特别设计了对潜在双重用途风险的处理，避免包含攻击性内容，并对潜在的不当请求提供解释性拒绝。

The Cybersecurity Defense Instruction-Tuning Dataset (v2.1) is a ready-to-train dataset of 99,870 high-quality system/user/assistant triples for defensive, alignment-safe cybersecurity supervised fine-tuning (SFT) training. It covers multiple cybersecurity standards and frameworks, including OWASP Top 10, MITRE ATT&CK, NIST CSF, CIS Controls, and focuses on cloud security, DevSecOps, modern authentication (OAuth 2/OIDC/SAML), SSL/TLS, cryptography, and AI security. The dataset is licensed under Apache-2.0 and is production-ready. The creation process involves source harvesting, extraction, topical filtering, instruction synthesis, and multi-layer quality gates to ensure high quality and safety. The dataset is designed to handle dual-use risks by avoiding malicious content and providing explanatory refusals for potential misuse requests.