Large-Scale Intrusion Detection Dataset (BCCC-CSE-CIC-IDS2018)

The BCCC-CSE-CIC-IDS2018 dataset is an enhanced version of CSE-CIC-IDS2018 with 46 million labelled records and 300 features, addressing key issues to improve data quality and reliability for behavioral profiling in IDS research. Labeling inconsistencies, particularly for DoS attacks, were corrected by aligning attack labels with attacker IPs instead of timestamps. NTLFlowLyzer, a new network traffic analyzer, was developed to resolve anomalies in extracted features and refine feature implementation. Additionally, protocol issues were fixed by removing UDP-based attacks previously misclassified due to TCP-specific analysis. Attacks with insufficient flow counts were retained but excluded from analysis and profiling. The dataset now includes an expanded feature set to detect evolving cyber threats better, making it a robust benchmark for AI-driven IDS/IPS research.