blackXmask/RedLockX-Prompt-Injection-109K-DataSet
收藏资源简介:
RedLockX数据集是一个大规模的安全数据集,专为评估和训练AI系统以对抗对抗性威胁而设计,包括提示注入、越狱尝试、系统提示泄露和LLM操纵攻击等。它包含了结构化真实世界和合成攻击模式,用于现代AI红队测试。数据集包含超过109,000个标记的对抗性和安全样本,具有多类别威胁分类系统、OWASP LLM十大威胁映射、严重性评分(0-10分制)和风险评分(0-100业务影响模型)。适用于微调、评估和基准测试,旨在提升AI安全性和对抗鲁棒性研究。
The RedLockX Dataset is a large-scale curated security dataset designed for evaluating and training AI systems against adversarial threats such as prompt injection, jailbreak attempts, system prompt leakage, and LLM manipulation attacks. It contains structured real-world and synthetic attack patterns used in modern AI red-teaming. The dataset includes 109,000+ labeled adversarial & safe samples, a multi-category threat classification system, OWASP LLM Top 10 mapping, severity scoring (0–10 scale), and risk scoring (0–100 business impact model). It is suitable for fine-tuning, evaluation, and benchmarking, aimed at enhancing AI safety and adversarial robustness research.




