Skip to content

Dataset mirror v1

Latest

Choose a tag to compare

@SackOfHacks SackOfHacks released this 19 Nov 04:08
· 7 commits to main since this release

Mirror of the ICS/OT/SCADA packet-capture datasets that their licences (or their authors) allow to be redistributed. Datasets without such a licence are not mirrored; scripts/fetch.py downloads those from their original sources.

Use it through the fetcher (it picks the mirror automatically, verifies every SHA-256 and unpacks):

python3 scripts/fetch.py --all          # or a dataset name, or --starter
python3 scripts/fetch.py tuwien-smart-factory --from-source   # force the original source instead

Format. One archive per dataset, split into parts under 2 GiB: <dataset>.tar.part000, .part001, ... Concatenate the parts in order and untar; packet captures inside are gzip-compressed (*.pcap.gz, restored byte-for-byte by gunzip). Checksums for every part are in catalog/datasets.json of the repository.

Licences. Every dataset keeps its own licence. Each archive contains LICENSE-NOTICE.md (authors, licence and link, why redistribution is allowed, original source, required citation) and, where the licence requires it, the original licence text. Several datasets are CC BY-NC / BY-NC-SA (non-commercial only) or CC BY-SA / GPL (share-alike). Credit the authors — see ATTRIBUTION.md.

Nothing in the packets was changed. The archives differ from the original downloads in three ways only: captures are gzip-compressed for transport, captures larger than 2 GB are split into < 2 GiB parts, and captures that were published without a file extension carry .pcap/.pcapng (chosen from their magic bytes).