CoolFace
Datasetpublic

lacg030175/CICIDS2017-raw

CICIDS2017 (raw variant) Companion to lacg030175/CICIDS2017. This variant preserves rows with NaN or ±infinity values in any feature column (the original dataset drops them via pd.dropna). Intended for use with ThermometerEncoder(invalid_encoding="single_bit"), which treats missing / undefined values as a learnable is_invalid flag bit rather than silently encoding them as zero. Row counts Full dataset (all days): 2,830,743 rows Splits: random (80/20): 2,264,594… See the full description on the dataset page: https://huggingface.co/datasets/lacg030175/CICIDS2017-raw.

sourceHugging Facecc-by-4.0updated 5mo agoView on Hugging Face
0likes54downloads
Dataset Card

CICIDS2017 (raw variant)

Companion to lacg030175/CICIDS2017. This variant preserves rows with NaN or ±infinity values in any feature column (the original dataset drops them via pd.dropna). Intended for use with ThermometerEncoder(invalid_encoding="single_bit"), which treats missing / undefined values as a learnable is_invalid flag bit rather than silently encoding them as zero.

Row counts

Full dataset (all days): 2,830,743 rows

Splits:

  • —random (80/20): 2,264,594 train / 566,149 test
  • —random_3way (80/10/10): 2,264,594 / 283,074 / 283,075
  • —temporal (Mon-Thu / Friday): 2,127,498 train / 703,245 test
  • —temporal_3way (train + Friday split 50/50): 2,127,498 / 351,622 / 351,623

Top-20 RF Features

  1. 1.Bwd Packet Length Std
  2. 2.Destination Port
  3. 3.Packet Length Std
  4. 4.Bwd Packet Length Max
  5. 5.Fwd IAT Std
  6. 6.Avg Bwd Segment Size
  7. 7.Bwd Packet Length Mean
  8. 8.Average Packet Size
  9. 9.Packet Length Mean
  10. 10.Flow IAT Max
  11. 11.Fwd IAT Max
  12. 12.Packet Length Variance
  13. 13.InitWinbytes_forward
  14. 14.Idle Mean
  15. 15.Idle Min
  16. 16.Flow IAT Std
  17. 17.Fwd Header Length
  18. 18.Flow Packets/s
  19. 19.Flow IAT Mean
  20. 20.Bwd Header Length

Labels

  • —Binary (label): 0 = BENIGN, 1 = Attack
  • —Multi-class (Label): 15 categories

Features

78 numeric flow-level features extracted by CICFlowMeter.

Preprocessing (raw variant)

  • —Rows with NaN / ±inf are PRESERVED (not dropped).
  • —Feature columns are still coerced to numeric via pd.to_numeric(errors="coerce"); unparseable strings become NaN.
  • —Label column preserved as multi-class string; binary label derived.
  • —Whitespace and encoding artifacts in Label are normalized.
  • —Use with ThermometerEncoder(invalid_encoding="single_bit") to encode NaN/±inf as a learnable state.

Citation

bibtex
@inproceedings{sharafaldin2018toward,
  title={Toward Generating a New Intrusion Detection Dataset and Intrusion Traffic Characterization},
  author={Sharafaldin, Iman and Lashkari, Arash Habibi and Ghorbani, Ali A},
  booktitle={International Conference on Information Systems Security and Privacy},
  year={2018}
}

License

CC BY 4.0 — original dataset by the Canadian Institute for Cybersecurity, University of New Brunswick.