lacg030175/CIC-IoT-2023-canonical-neto

这是规范的CIC-IoT-2023数据集，源自Neto等人的真实合并CSV文件，包含嵌入式标签。与之前版本相比，这个版本保留了所有约45,019,243行数据（Neto的原始计数），并且没有删除NaN或±Inf值。数据集提供了原始标签（`Label_orig`）以及标准化后的标签（`Label`、`attack_class`和二进制标签`label`）。数据分割包括`random_3way`（80%训练，10%测试，10%验证）和`random`（80%训练，20%测试）。类别分布显示，DDoS攻击占72.27%，DoS攻击占17.21%，其他类别如Mirai、Recon、Spoofing等占比较小。

This is the **canonical** CIC-IoT-2023 dataset, sourced from Neto et al.s authentic merged CSVs WITH embedded labels. Compared to previous versions, this one contains **~45,019,243 rows** (the actual canonical Neto count) and preserves NaN or ±Inf values. The dataset provides original labels (`Label_orig`) and normalized labels (`Label`, `attack_class`, and binary `label`). Splits include `random_3way` (80% train, 10% test, 10% validation) and `random` (80% train, 20% test). The class distribution shows DDoS attacks account for 72.27%, DoS for 17.21%, and other categories like Mirai, Recon, Spoofing, etc., for smaller percentages.