nguynn/cert-insider-threat-datasets

CERT Insider Threat & Security Log Datasets

Mirror of several public cybersecurity research datasets, organized by source for easier reuse. Each folder keeps its own license/readme files - check them before redistributing.

Folder structure

Folder	What it is	License
cert_r4.2/	CERT r4.2 synthetic insider-threat dataset (extracted CSVs: logon.csv, device.csv, http.csv, email.csv, file.csv, psychometric.csv, LDAP/).	see cert_r4.2/license.txt
cmu_kilthub/	CMU KiltHub "Insider Threat Test Dataset" releases r1 ... r6.2 (extracted CSVs; r*/r*.tar.bz2 archives may also be present), plus answers/ (ground-truth malicious-actor annotations) and docs/. Source: https://doi.org/10.1184/R1/12841247.v1	CC BY 4.0
kaggle/	Mirror of Kaggle mrajaxnp/cert-insider-threat-detection-research.	see source
kaggle_emmanuelkiriinya_r6.2/	Mirror of Kaggle emmanuelkiriinya/insider-threat-dataset-r6-2. File names, sizes and SHA256 are identical to cmu_kilthub/r6.2/.	same as CMU r6.2
otrf_security_datasets/	Clone of OTRF/Security-Datasets (Mordor): JSON-lines security event logs for simulated attacks (Windows/Linux/AWS, atomic + compound such as APT29/APT3). Original .zip/.tar.gz are kept next to the extracted files. APT29 pcap zips use password infected.	MIT (see otrf_security_datasets/LICENSE)
splunk_bots/	Splunk Boss of the SOC botsv1, botsv2, botsv3 (from the Splunk GitHub repos + the S3 downloads they link to). Each botsvN/ has README.md, LICENSE and full/ (and attack_only/ for v1/v2) with the extracted pre-indexed Splunk data.	see each botsvN/LICENSE

Use in a notebook

See notebooks/load_dataset_example.ipynb. Quick start:

%pip install -q -U huggingface_hub pandas pyarrow
import pandas as pd
from huggingface_hub import snapshot_download

REPO = "nguynn/cert-insider-threat-datasets"

# read a small file directly from the Hub
psy = pd.read_csv(f"hf://datasets/{REPO}/cmu_kilthub/r4.2/psychometric.csv")

# download only the files you need (CMU releases are 10-100 GB each)
local = snapshot_download(REPO, repo_type="dataset",
                          allow_patterns=["cmu_kilthub/r4.2/logon.csv", "cmu_kilthub/r4.2/LDAP/*"])

Notes:

• CERT CSVs: dates are MM/DD/YYYY HH:MM:SS; http.csv/email.csv are multi-GB, read with chunksize.
• OTRF logs are JSON-lines: pd.read_json(path, lines=True).
• BOTS data is pre-indexed Splunk data (not CSV/JSON); load it into a Splunk instance as described in each botsvN/README.md.

Sources

• CMU KiltHub: Lindauer, Brian (2020). Insider Threat Test Dataset. Carnegie Mellon University. https://doi.org/10.1184/R1/12841247.v1
• Kaggle mirrors: see links in the table above.
• OTRF Security-Datasets: https://github.com/OTRF/Security-Datasets
• Splunk BOTS: https://github.com/splunk/botsv1, https://github.com/splunk/botsv2, https://github.com/splunk/botsv3

The CERT/CMU datasets are synthetic (no real user data).