README.md

---
license: other
license_name: cic-ids-2017-research-use
license_link: https://www.unb.ca/cic/datasets/ids-2017.html
task_categories:
- tabular-classification
language:
- en
tags:
- tsfile
- time-series
- network-security
- cybersecurity
- netflow
- flow
- intrusion-detection
- nids
- canonical-schema
- flowprep
- deeptempo
pretty_name: CIC-IDS-2017 Canonical NetFlow Flowprep TsFile
size_categories:
- 100K<n<1M
modality: timeseries
configs:
- config_name: default
  data_files:
  - split: train
    path: cic_ids_2017_flowprep.tsfile
---

# CIC-IDS-2017 Canonical NetFlow Flowprep (TsFile)

This repository contains an Apache TsFile conversion of
[`DeepTempo/cic-ids-2017-flowprep`](https://huggingface.co/datasets/DeepTempo/cic-ids-2017-flowprep),
a small CIC-IDS-2017 demonstration slice canonicalized by DeepTempo's
`flowprep` tool into a typed NetFlow schema.

Modalities: Time-series.

## Source Dataset

- **Original dataset:** [`DeepTempo/cic-ids-2017-flowprep`](https://huggingface.co/datasets/DeepTempo/cic-ids-2017-flowprep)
- **Source artifact:** `data/cic-ids-2017-canonical.parquet`
- **Rows:** 101,094 flows
- **Columns:** 13 source columns
- **Task:** binary intrusion-detection / tabular classification
- **Source format:** ZSTD-compressed Parquet, single row group
- **Source timestamp encoding:** int64 epoch microseconds
- **Source license metadata:** `other`, `cic-ids-2017-research-use`
- **License link:** https://www.unb.ca/cic/datasets/ids-2017.html

The source dataset is a clean canonical NetFlow table produced by `flowprep`
from a CIC-IDS-2017 sample. It is a demonstration slice, not the full
CIC-IDS-2017 dataset. For research use, refer to the official UNB CIC dataset
page and cite the original CIC-IDS-2017 paper.

## Converted Data

- **TsFile path:** `cic_ids_2017_flowprep.tsfile`
- **TsFile table:** `cic_ids_2017_flowprep`
- **Rows:** 101,094
- **Converted columns:** 14 including `Time` and generated `event_rank`
- **Device/TAG groups:** 1,780
- **Time precision:** microseconds
- **Time range:** 2017-03-07 01:00:01 UTC to 2017-07-07 12:59:00 UTC
- **Class balance:** 81,171 benign / 19,923 attack

## TsFile Schema

`timestamp` is converted to the TsFile `Time` column as epoch microseconds and
is not retained as a duplicate FIELD.

TAG columns:

- `attack`
- `label`
- `event_rank`

FIELD columns:

- `src_ip`
- `dest_ip`
- `src_port`
- `dest_port`
- `fwd_bytes`
- `bwd_bytes`
- `fwd_pkts`
- `bwd_pkts`
- `flow_dur`
- `protocol`

`protocol` is null for all rows in this source slice and is preserved as a
nullable numeric FIELD for canonical-schema fidelity.

## Conversion Notes

This is a network-flow event table. High-cardinality endpoint columns such as
`src_ip`, `dest_ip`, `src_port`, and `dest_port` are kept as FIELD columns rather
than TAG/device keys. The low-cardinality ground-truth columns `attack` and
`label` are TAGs for efficient filtering.

`event_rank` is a generated TAG that preserves all concurrent flows without
modifying `Time`. It is the duplicate order within `(attack, label, Time)`.
In this source snapshot, `event_rank` ranges from 0 to 1,587 and 88,916 rows
have a nonzero rank. The final `(attack, label, event_rank, Time)` key has no
duplicates.

No source rows are dropped. The source `timestamp` column is represented by
TsFile `Time`; all other source columns are represented either as TAG or FIELD
columns.

## Minimal Read Example

Read the `.tsfile` file with the Apache TsFile Java or Python SDK.

Example logical filter:

```sql
SELECT *
FROM cic_ids_2017_flowprep
WHERE attack = 'attack'
```

## Citation

If you use the data, cite the original CIC-IDS-2017 paper:

```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 = {Proceedings of the 4th International Conference on Information Systems Security and Privacy (ICISSP)},
  year      = {2018}
}
```