anonymizeddb/CAD-CICUNSW

CAD-CICUNSW 是一个用于网络入侵检测的单源持续异常检测基准场景。它源自 CIC-UNSW-NB15/UNSW-NB15 数据集，将原始的表格型网络入侵数据转换为按概念分组的任务序列。数据集包含 1,084,928 个样本和 5 个任务，测试集中报告的异常比率为 12.76%。该数据集已匿名化，适用于双盲NeurIPS评审，旨在支持防御性机器学习研究，包括持续异常检测、表格数据的持续学习、网络入侵检测、分布偏移下的鲁棒性、持续学习基准中的任务排序、跨相关网络流量概念的遗忘和知识转移，以及在顺序任务暴露下的异常检测器基准测试。数据集文件包括 data.csv（主表格文件）、orderings.json（预定义任务排序）和 croissant.json（元数据）。核心列包括任务ID、任务名称、任务拆分标签和二进制异常标签（0表示良性/正常流量，1表示异常/攻击流量），特征列涵盖数值型网络流特征，如源端口、目标端口、协议、流持续时间、数据包计数、字节计数等。数据集提供六种预定义任务排序，以评估不同的持续学习动态，如课程式适应、泛化导向排序、平滑漂移和突然漂移。CAD-CICUNSW is a single-source continual anomaly detection benchmark scenario for network intrusion detection. It is derived from CIC-UNSW-NB15/UNSW-NB15 and converts the original tabular network-intrusion data into a sequence of concept-grouped tasks. The dataset contains 1,084,928 samples, 5 tasks, and has a reported 12.76% anomaly ratio in the test set. The dataset is anonymized for double-blind NeurIPS review and intended for defensive machine learning research, including continual anomaly detection, continual learning for tabular data, network intrusion detection, robustness under distribution shift, task ordering in continual-learning benchmarks, forgetting and knowledge transfer across related network-traffic concepts, and benchmarking anomaly detectors under sequential task exposure. Dataset files include data.csv (main tabular file), orderings.json (predefined task orderings), and croissant.json (metadata). Core columns include task ID, task name, task split label, and binary anomaly label (0 for benign/normal traffic, 1 for anomalous/attack traffic), with feature columns covering numerical network-flow features such as source port, destination port, protocol, flow duration, packet counts, byte counts, etc. The dataset provides six predefined task orderings to expose complementary continual-learning dynamics like curriculum-like adaptation, generalization-oriented ordering, smooth drift, and abrupt drift.