flwrlabs/fed-phishing-urls
收藏资源简介:
该数据集是一个联邦学习、非独立同分布(non-IID)钓鱼URL分类基准数据集,来源于两个公开的Hugging Face数据集。数据集包含URL字符串、二进制钓鱼标签(1表示钓鱼或恶意,0表示良性或合法)和模拟客户端ID字段(范围0到99)。数据集经过预处理,包括URL字符串和标签的合并、去重、特征提取、客户端分配和分割等步骤。数据集分为训练集和测试集,适用于联邦二元分类、跨客户端泛化研究、非IID鲁棒性基准测试、钓鱼URL检测、网络安全模型评估等用途。
This dataset is a benchmark dataset for federated learning and non-independent and identically distributed (non-IID) phishing URL classification, sourced from two public Hugging Face datasets. It contains URL strings, binary phishing labels (1 indicates phishing or malicious, 0 indicates benign or legitimate), and simulated client ID fields ranging from 0 to 99. The dataset has undergone preprocessing steps including merging URL strings and their corresponding labels, deduplication, feature extraction, client assignment, and dataset splitting. The dataset is split into training and test sets, and is suitable for applications such as federated binary classification, cross-client generalization research, non-IID robustness benchmarking, phishing URL detection, and cybersecurity model evaluation.




