Hugging Face's logo

Datasets:
TocharianOU
/
secops-es-benchmark

Tasks:
Question Answering
Text Generation
Languages:
English
Size:
100K<n<1M
Tags:
security
cybersecurity
secops
siem
elasticsearch
ecs
License:
Dataset card Files
xet
 Community
secops-es-benchmark / ETHICS.md
azmatjan139's picture
azmatjan139
Upload 87 files
346127e verified
preview code
|
Raw
History Blame Contribute Delete
2.38 kB

Ethics, Legal & Responsible-Use Statement

Authorization & ownership

All systems used to generate this dataset — the attacker/C2 host, both victim hosts, and the Elasticsearch cluster — are owned and operated by the dataset authors. All attack activity was self-authorized red-teaming on our own infrastructure. No third-party systems were scanned, accessed, or attacked. No unauthorized access occurred.

No third-party or personal data

  • The environment is a controlled lab/production-of-our-own; it does not contain customer PII by design. Any staged "sensitive" data in the attack scenarios is synthetic decoy (e.g., DECOY_AKIA_BENCH, fabricated CSV rows).
  • Secrets and business/PII are removed before release (see DATASHEET.md and benchmark/lib/pseudonymize.py): a DB password, OS password hashes, private keys, API tokens, email addresses, and business identifiers do not appear in the distributed files (verified 0 residual). Network identifiers (IPs, hostnames) appear as captured so the data stays investigable; third-party addresses in the background noise are public-actor network metadata (scanners, public services), retained as-is.

Safety of the artifacts

  • The dataset is log data only. Strings like /tmp/.sysupdate, /tmp/.shell.php, or reverse-shell command lines are inert text in event records — no malware binaries, no live C2, no exploit payloads are distributed.
  • The included attack scripts (corpus/scenarios/*.sh) perform non-destructive, well-known ATT&CK techniques for detection research. They require a Sliver C2 and a consenting lab to run and are provided for reproducibility, not for use against systems you do not own.

Intended use & prohibited use

  • Intended: benchmarking SOC/AI investigation agents, detection engineering, teaching, and correlation research.
  • Prohibited: using the scenarios or any derived tooling against systems you are not authorized to test; attempting to re-identify the source environment.

Licensing

Data under CC-BY-4.0 (LICENSE-DATA); code/scripts under Apache-2.0 (LICENSE-CODE). Attribution appreciated per the dataset citation in README.md.

Contact / disclosure

Issues, re-identification concerns, or takedown requests: open an issue at https://github.com/TocharianOU/secops-es-benchmark/issues