Zieni dataset for Phishing detection

This dataset was used for training machine learning models to detect phishing attacks and for studying the explainability of these models. It was published in 2024. The dataset refers to phishing and legitimate websites. Phishing samples have been collected from two sources, namely, PhishTank and Tranco, whereas legitimate samples were collected from Alexa. The dataset is balanced and contains 5,000 phishing and 5,000 legitimate samples, each described by 74 features extracted from the entire URL as well as from the Fully Qualified Domain Name, pathname, filename, and parameters. Of these features, 70 are numerical and four binary. The target variable is also binary.本数据集用于训练网络钓鱼攻击检测机器学习模型，同时可用于相关模型的可解释性研究，于2024年正式发布。该数据集涵盖网络钓鱼与合法网站两类样本：网络钓鱼样本采集自PhishTank与Tranco两个数据源，合法样本则来源于Alexa。数据集样本类别均衡，共包含5000个网络钓鱼样本与5000个合法样本，每个样本均通过74项特征进行表征，这些特征提取自完整统一资源定位符(URL)、完全限定域名(Fully Qualified Domain Name)、路径名、文件名及参数。其中70项为数值型特征，剩余4项为二分类特征，目标变量同样为二分类类型。