Acti
收藏资源简介:
Acti数据集由北京航空航天大学创建,专注于挖掘自动驾驶车辆网络安全威胁情报的实体及其关联。该数据集包含908份真实的汽车网络安全报告,涵盖3678个句子、8195个安全实体和4852个语义关系。数据集的创建过程包括从国家漏洞数据库和特定车辆威胁情报平台收集数据,并通过BIOES联合标注策略进行标注。Acti数据集主要应用于汽车网络安全威胁情报建模,旨在通过知识提取技术从大量网络安全数据中获取有价值信息,以实现主动安全防御。
The Acti dataset was developed by Beihang University, focusing on extracting entities and their relationships associated with cybersecurity threat intelligence for autonomous vehicles. This dataset comprises 908 real automotive cybersecurity reports, including 3678 sentences, 8195 security entities, and 4852 semantic relations. The dataset construction process involves collecting data from the National Vulnerability Database (NVD) and specialized vehicle threat intelligence platforms, followed by annotation using the BIOES joint annotation scheme. The Acti dataset is primarily utilized for automotive cybersecurity threat intelligence modeling, with the objective of extracting valuable information from large-scale cybersecurity data via knowledge extraction technologies to enable proactive security defense.
Automotive-cyber-threat-intelligence-corpus
数据集概述
该数据集用于连接自动驾驶车辆的网络威胁情报建模。
实验环境
- NVIDIA GeForce RTX 3090 GPU
- Python 3.7
- CUDA 11.2
- PaddlePaddle-GPU 2.3.2
- paddlenlp 2.1.1
数据描述
- 原始数据: 非结构化的网络安全数据(.txt文件)
- Brat标注数据: 使用brat工具的标注数据文件(.ann文件)
- BIOES: "BIOES" - "实体类型" - "关系类型" - "实体角色" 联合标注数据(.txt文件)
源代码描述
- 格式转换: BIOES联合标注.py
- 预处理: read.py; preprocess.py
- 深度学习模型训练: BERT-BiLSTM-att-CRF; BiLSTM-dynamic-att-LSTM
Brat工具
- https://github.com/nlplab/brat/archive/refs/tags/v1.3p1.tar.gz




