lacg030175/CICIDS2017-raw

CICIDS2017 (原始变体)是一个网络入侵检测数据集，主要用于表格分类任务。该数据集保留了包含NaN或无限值的行，适用于特定的编码器处理。数据集包含2,830,743行数据，分为不同的训练、测试和验证集。数据集提供了78个数字流级特征，标签分为二分类（0 = 良性，1 = 攻击）和多分类（15个类别）。数据预处理保留了NaN/±inf值，标签列进行了规范化处理。数据集由加拿大网络安全研究所创建，采用CC BY 4.0许可。

CICIDS2017 (raw variant) is a network intrusion detection dataset primarily used for tabular classification tasks. This variant preserves rows with NaN or infinity values, making it suitable for specific encoder processing. The dataset contains 2,830,743 rows of data, divided into different training, testing, and validation sets. It provides 78 numeric flow-level features, with labels categorized into binary (0 = BENIGN, 1 = Attack) and multi-class (15 categories). Preprocessing preserves NaN/±inf values, and the label column is normalized. The dataset was created by the Canadian Institute for Cybersecurity and is licensed under CC BY 4.0.