Licensing and provenance caveats
This release is intentionally conservative. It does not claim to relicense upstream corpora, and it does not assume that a public mirror automatically authorizes downstream public redistribution.
High-level corpus posture
| Corpus | Public CSVs shipped here? | Current reading |
|---|---|---|
| HIKARI-2021 | Yes | Strong documented basis via CC BY 4.0 route |
| 5G-NIDD | Yes | Strong documented basis via open Fairdata CC BY 4.0 route |
| RT-IoT2022 | Yes | Strong documented basis via UCI CC BY 4.0 route |
| CIC-IDS-2017 | Yes | Included with CIC provenance caveats; redistribution language appears supportive, but derivative-subset rights are not framed as a standard open-data license |
| CSE-CIC-IDS2018 | Yes | Included with CIC provenance caveats; redistribution language appears broad, but this repo does not treat that as a blanket relicensing |
| Edge-IIoTset | No | Public mirror terms appear noncommercial/share-alike; omitted from the public bundle to avoid overclaiming rights |
| CIC-UNSW-NB15 | No | Academic use is explicit, but redistribution of repackaged benchmark subsets was not explicitly verified |
| CICIoMT2024Small mirror | No | Mirror-derived small artifact was the benchmark input, but mirror-specific license terms were unclear |
Where status is recorded per task
Every task row in metadata/datasets.csv and metadata/datasets.json includes downloadable, public_bundle_status, public_csv, omission_reason, omission_caveat, and redistribution_caveat. The downloadable task catalog repeats the same status next to each direct CSV link.
Downloadable means the public release includes a CSV under data/. Omitted/not redistributed means no public CSV is shipped here; the task is retained only as metadata plus a local/private recreation workflow.
Why mirrors matter
A public mirror can be useful for reproducibility, but it does not automatically prove that every downstream redistribution is authorized. This matters most when the benchmark used a mirrored artifact instead of an official open release or when the official route is gated and the public mirror carries separate terms.