s0u9ata/security-kg

Security Knowledge Graph Triples

Security data from 17 sources represented as Subject-Predicate-Object (SPO) triples in Parquet format, ready for knowledge-graph construction, graph-ML, RAG pipelines, and threat-intelligence analysis.

Sources: ATT&CK · CAPEC · CWE · CVE · CPE · D3FEND · ATLAS · CAR · ENGAGE · F3 · EPSS · KEV · Vulnrichment · GHSA · Sigma · ExploitDB · MISP Galaxies

Last updated: 2026-04-27T08:30:00Z

Quick Start

from datasets import load_dataset

ds = load_dataset("s0u9ata/security-kg", "enterprise")
print(ds["train"][0])
# {'subject': 'T1059.001', 'predicate': 'rdf:type', 'object': 'Technique', 'source': 'attack', 'object_type': 'enum', 'meta': ''}

Configurations

Config	Description	Est. Triples	Status
enterprise (default)	Enterprise ATT&CK	42,041	Current
mobile	Mobile ATT&CK	5,307	Current
ics	ICS ATT&CK	3,756	Current
attack-all	ATT&CK combined (deduplicated)	49,622	Current
capec	CAPEC attack patterns	8,114	Current
cwe	CWE weaknesses	14,565	Current
cve	CVE vulnerabilities	3,622,414	Current
cpe	CPE platform enumeration	12,669,910	Current
d3fend	D3FEND defensive techniques	8,154	Current
atlas	ATLAS AI/ML techniques	1,420	Current
car	CAR analytics	1,617	Current
engage	ENGAGE adversary engagement	1,464	Current
f3	F3 fraud techniques & tactics	1,047	Current
epss	EPSS exploit prediction scores	658,266	Current
kev	KEV known exploited vulns	17,343	Current
vulnrichment	CISA Vulnrichment (SSVC, CVSS, CWE enrichment)	670,832	Current
ghsa	GitHub Security Advisories	338,240	Current
sigma	Sigma detection rules	32,750	Current
exploitdb	ExploitDB public exploits	346,471	Current
misp_galaxy	MISP Galaxy threat intelligence clusters	196,489	Current
combined	All sources merged (deduplicated)	18,638,718	Current

Knowledge Graph Structure

     Group  Campaign
        \      /
          uses
           |
           v
      TECHNIQUE -----> Tactic
        ^  ^  ^
        |  |  |
        |  |  +-- D3FEND (counters)
        |  |  +-- CAR (detects)
        |  |  +-- Sigma (detects)
        |  |  +-- ENGAGE (engages)
        |  |  +-- F3 (fraud techniques)
        |  |  +-- ATLAS (related)
        |  |  +-- MISP Galaxies (cross-refs)
        |  |
        |  +-- Mitigation (mitigates)
        |  +-- DataComponent (detects)
        |
        +-- maps-to -- CAPEC
                         |
                  related-weakness
                         |
                         v
                        CWE
                         ^
                         |
                  related-weakness
                         |
                        CVE ----> CPE
                         ^
                         |
                   EPSS (score)
                   KEV (exploited)
                   GHSA (advisory)
                   Vulnrichment (SSVC)
                   ExploitDB (exploit)

Schema

Each row is an enriched triple with six string columns:

Column	Description	Examples
subject	Entity ID	T1059.001, G0016, CAPEC-66, CWE-79, CVE-2024-1234, cpe:2.3:a:apache:httpd:*, D3-FE, AML.T0000, CAR-2024-01-001, EAC0001, GHSA-xxxx-yyyy-zzzz, EDB-16929
predicate	Property name or relationship type	rdf:type, name, uses, mitigates, epss-score, counters, ssvc-exploitation, exploits-cve, detects-technique
object	Value or target entity ID	Technique, PowerShell, T1059, CWE-89, 0.97500, SecurityAdvisory, SigmaRule, Exploit
source	Originating dataset	attack, cve, cwe, capec, epss, kev, ghsa, sigma, d3fend, atlas, car, engage, f3, cpe, vulnrichment, exploitdb, misp_galaxy
object_type	Value type of the object	string, id, enum, date, number, boolean, url
meta	Supplemental JSON metadata (empty string if none)	{"references":["https://..."],"credits":[...]}, {"cvss_vector":"...","cvss_version":"3.1"}

Predicate Reference

ATT&CK Entity Properties

Predicate	Description	Example object value
rdf:type	Entity type	Technique, Group, Malware, Tool, Tactic, Mitigation, Campaign, DataSource, DataComponent
name	Display name	PowerShell
description	Full description text	Adversaries may abuse PowerShell...
platform	Applicable platform	Windows, Linux, macOS
domain	ATT&CK domain	enterprise-attack
alias	Alternative name	Cozy Bear
is-subtechnique	Whether entity is a sub-technique	True, False
belongs-to-tactic	Tactic ATT&CK ID	TA0002
shortname	Tactic shortname	credential-access
url	ATT&CK website URL	https://attack.mitre.org/techniques/T1059/001
created / modified	Timestamps	2020-01-14 17:18:32...

ATT&CK Relationship Predicates

Predicate	Typical subject / object	Example
uses	Group/Campaign/Software / Technique	G0016 / T1059.001
mitigates	Mitigation / Technique	M1049 / T1059.001
subtechnique-of	Sub-technique / Parent technique	T1059.001 / T1059
detects	DataComponent / Technique	DC0001 / T1059.001
attributed-to	Campaign / Group	C0018 / G0016

CAPEC Predicates

Predicate	Description	Example object value
rdf:type	AttackPattern	AttackPattern
name / description	Display name / full text	SQL Injection
abstraction / status	Level / status	Standard, Stable
likelihood / severity	Attack likelihood / severity	High
child-of	Parent attack pattern	CAPEC-248
related-weakness	Related CWE	CWE-89
maps-to-technique	Mapped ATT&CK technique	T1190.002

CWE Predicates

Predicate	Description	Example object value
rdf:type	Weakness	Weakness
name / description	Display name / full text	Cross-site Scripting (XSS)
abstraction / status	Level / status	Base, Stable
likelihood-of-exploit	Exploitation likelihood	High
child-of	Parent weakness	CWE-74
related-attack-pattern	Related CAPEC	CAPEC-86
platform	Applicable platform	JavaScript
consequence-scope / consequence-impact	Impact	Confidentiality, Read Data
introduction-phase	Introduction phase	Implementation

CVE Predicates

Predicate	Description	Example object value
rdf:type	Vulnerability	Vulnerability
state	CVE state	PUBLISHED
description	English description	A remote code execution...
date-published / date-updated	Timestamps	2024-01-15T00:00:00.000Z
assigner	Assigning organization	microsoft
vendor / product	Affected vendor/product	Microsoft, Windows
affects-cpe	Affected CPE string	cpe:2.3:o:microsoft:windows_10:*
platform	Affected platform	x64
related-weakness	Related CWE	CWE-79
cvss-base-score / cvss-severity	CVSS metrics	9.8, CRITICAL

CPE Predicates

Predicate	Description	Example object value
rdf:type	Platform	Platform
part	CPE part type	application, operating_system, hardware
vendor / product / version	Components	apache, httpd, 2.4.51
title	English display name	Apache HTTP Server 2.4.51
created / modified	Timestamps	2021-10-07

D3FEND Predicates

Predicate	Description	Example object value
rdf:type	DefensiveTechnique or OffensiveTechnique	DefensiveTechnique
name / definition	Display name / definition	File Encryption
synonym	Alternative name	Disk Encryption
child-of	Parent technique	PlatformHardening
counters	Countered offensive technique	T1059

ATLAS Predicates

Predicate	Description	Example object value
rdf:type	Tactic, Technique, CaseStudy, Mitigation	Technique
name / description	Display name / full text	ML Supply Chain Compromise
maturity	Technique maturity	Reviewed
belongs-to-tactic	Parent tactic	AML.TA0001
subtechnique-of	Parent technique	AML.T0000
related-attack-technique	Linked ATT&CK technique	T1195
related-attack-tactic	Linked ATT&CK tactic	TA0001
uses-technique	Case study technique	AML.T0000
mitigates	Mitigated technique	AML.T0000

CAR Predicates

Predicate	Description	Example object value
rdf:type	Analytic	Analytic
title / description	Analytic name / full text	Suspicious PowerShell Commands
platform	Applicable platform	Windows
information-domain	Information domain	Host
analytic-type	Type of analytic	Situational Awareness
detects-technique	Detected ATT&CK technique	T1059
detects-subtechnique	Detected subtechnique	T1059.001
covers-tactic	Covered ATT&CK tactic	Execution
maps-to-d3fend	Linked D3FEND technique	D3-PSA

ENGAGE Predicates

Predicate	Description	Example object value
rdf:type	EngagementActivity or AdversaryVulnerability	EngagementActivity
name / description	Display name / full text	Software Manipulation
engages-technique	Engaged ATT&CK technique	T1001
vulnerability-of	ATT&CK technique this adversary vulnerability applies to	T1001
addresses-vulnerability	Addressed adversary vulnerability	EAV0001

F3 Predicates

Predicate	Description	Example object value
rdf:type	Tactic or Technique	Technique
name / description	Display name / full text	Account Takeover
shortname	Tactic shortname	positioning, monetization
is-subtechnique	Whether entity is a sub-technique	true
belongs-to-tactic	Parent tactic	FA0001
subtechnique-of	Parent technique	F1001
url	F3 website URL	https://ctid.mitre.org/fraud/techniques/F1001
created / modified	Timestamps	2026-04-02T19:15:57.686Z

EPSS Predicates

Predicate	Description	Example object value
epss-score	Exploit probability (0-1)	0.97500
epss-percentile	Score percentile (0-1)	0.99900

KEV Predicates

Predicate	Description	Example object value
rdf:type	KnownExploitedVulnerability	KnownExploitedVulnerability
kev-vendor / kev-product	Affected vendor/product	Microsoft, Windows
kev-name / kev-description	Vulnerability name/description	Windows Privilege Escalation
kev-date-added / kev-due-date	Dates	2024-01-15
kev-required-action	Required remediation action	Apply updates per vendor instructions.
kev-ransomware-use	Ransomware campaign use	Known, Unknown
related-weakness	Related CWE	CWE-269

Vulnrichment Predicates

Predicate	Description	Example object value
ssvc-exploitation	SSVC exploitation status	active, poc, none
ssvc-automatable	Whether exploitation is automatable	yes, no
ssvc-technical-impact	Technical impact level	total, partial
adp-cvss-base-score	CISA-analyzed CVSS base score	9.8
adp-cvss-severity	CISA-analyzed CVSS severity	CRITICAL
adp-related-weakness	CISA-assigned CWE	CWE-79
adp-affects-cpe	CISA-assigned CPE	cpe:2.3:o:microsoft:windows_10:*

GHSA Predicates

Predicate	Description	Example object value
rdf:type	SecurityAdvisory	SecurityAdvisory
summary	Advisory summary	XSS vulnerability in example-package
date-published / date-modified	Timestamps	2024-01-15T00:00:00Z
severity	Severity level	HIGH, MODERATE, LOW, CRITICAL
related-cve	Associated CVE	CVE-2024-1234
related-weakness	Associated CWE	CWE-79
cvss-vector	CVSS v3 vector string	CVSS:3.1/AV:N/AC:L/...
affects-package	Affected package (ecosystem/name)	npm/example-package
fixed-in	Fixed version for package (ecosystem/name@version)	npm/example-package@2.0.1

Sigma Predicates

Predicate	Description	Example object value
rdf:type	SigmaRule	SigmaRule
title / description	Rule name / full text	Suspicious PowerShell Download
status	Rule maturity	stable, test, experimental
level	Detection severity	critical, high, medium, low, informational
author / date	Rule author / creation date	Security Researcher, 2024-01-15
logsource-category	Log source category	process_creation, network_connection
logsource-product	Log source product	windows, linux
logsource-service	Log source service	sshd, sysmon
detects-technique	Detected ATT&CK technique	T1059.001
related-cve	Related CVE	CVE-2024-1234

ExploitDB Predicates

Predicate	Description	Example object value
rdf:type	Exploit	Exploit
description	Exploit description	Apache HTTP Server RCE
date-published	Publication date	2024-01-15
author	Exploit author	Metasploit
exploit-type	Exploit category	remote, local, dos, webapps
platform	Target platform	linux, windows, aix
verified	Verified by OffSec	True
exploits-cve	Exploited CVE	CVE-2024-1234

MISP Galaxy Predicates

Predicate	Description	Example object value
rdf:type	Galaxy entity type	ThreatActor, Ransomware, Botnet, RAT
name	Display name	APT1
description	Full description	(text)
galaxy	Galaxy cluster type	threat-actor, ransomware
synonym	Alternative name	Comment Crew
country	Country code (ISO 3166-1)	CN
cfr-suspected-state-sponsor	Suspected state sponsor	China
targets-country	Targeted country	United States
targets-sector	Targeted sector	Government
attribution-confidence	Confidence level	50
similar-to	Similar/duplicate entity	misp:<uuid>
uses	Uses technique/tool	misp:<uuid>
used-by	Used by actor	misp:<uuid>
variant-of	Variant relationship	misp:<uuid>
targets	Targets entity	misp:<uuid>
attributed-to	Attributed to entity	misp:<uuid>
misp-related	Generic relationship	misp:<uuid>
related-attack-id	Cross-link to ATT&CK	T1059.001, G0006

Dataset Creation

Source Data

Source	Feed	Format
ATT&CK	mitre-attack/attack-stix-data	STIX 2.0 JSON
CAPEC	capec_latest.xml	XML
CWE	cwec_latest.xml.zip	XML (ZIP)
CVE	CVEProject/cvelistV5	JSON 5.x (ZIP)
CPE	nvdcpe-2.0.tar.gz	JSON (tar.gz)
D3FEND	d3fend.json	JSON-LD
ATLAS	ATLAS.yaml	YAML
CAR	mitre-attack/car	YAML (ZIP)
ENGAGE	attack_mapping.json	JSON
F3	fight-fraud-framework	STIX 2.1 JSON
EPSS	epss_scores-current.csv.gz	CSV (gzip)
KEV	known_exploited_vulnerabilities.json	JSON
Vulnrichment	cisagov/vulnrichment	JSON 5.x (ZIP)
GHSA	github/advisory-database	OSV JSON (ZIP)
Sigma	SigmaHQ/sigma	YAML (ZIP)
ExploitDB	files_exploits.csv	CSV
MISP Galaxies	MISP/misp-galaxy	JSON (ZIP)

Conversion Pipeline

The converter downloads source data, extracts entity property triples and relationship triples, and writes them as Parquet files. The source code and full documentation are at:

github.com/S0UGATA/security-kg

To regenerate or update this dataset:

git clone https://github.com/S0UGATA/security-kg.git
cd security-kg
pip install -r requirements.txt
python src/convert.py

This produces fresh Parquet files in output/ from the latest data across all 17 sources.

Visualizer

Explore the Parquet files interactively at security-kg-viz.

Use Cases

• Knowledge Graph Construction: Load triples into Neo4j, RDFLib, or NetworkX for graph queries
• Graph ML: Train graph neural networks (GNNs) on security data structure for link prediction
• RAG / LLM Grounding: Use triples as structured context for retrieval-augmented generation
• Threat Intelligence: Query relationships between groups, techniques, vulnerabilities, and mitigations
• Vulnerability Prioritization: Combine SSVC, EPSS, KEV, and ExploitDB data for risk-based triage
• Defensive Gap Analysis: Find heavily-used ATT&CK techniques with insufficient detection coverage
• Supply Chain Risk: Score open-source packages by linking GHSA advisories to CVE/EPSS/KEV enrichment
• Security Automation: Programmatically map detections to techniques to tactics

Cross-Source Analysis Notebook

The repository includes a Jupyter notebook with 16 cross-source analyses and visualizations built on combined.parquet — covering SSVC patch prioritization, defensive gap analysis, kill chain tactic coverage, exploit weaponization timelines, ransomware CWE pipelines, supply chain package risk, and more.

Example Queries

SSVC Patch Prioritization (Vulnrichment + EPSS + KEV)

import pandas as pd
from datasets import load_dataset

# Load combined graph for cross-source queries
ds = load_dataset("s0u9ata/security-kg", "combined")
df = ds["train"].to_pandas()

# Build SSVC triage matrix: exploitation status × automatable × EPSS score
ssvc = df[df.predicate == "ssvc-exploitation"][["subject", "object"]].rename(columns={"object": "exploitation"})
auto = df[df.predicate == "ssvc-automatable"][["subject", "object"]].rename(columns={"object": "automatable"})
epss = df[df.predicate == "epss-score"][["subject", "object"]].copy()
epss["epss"] = epss.object.astype(float)

triage = ssvc.merge(auto, on="subject").merge(epss[["subject", "epss"]], on="subject")

# Highest priority: actively exploited + automatable + high EPSS
critical = triage[(triage.exploitation == "active") & (triage.automatable == "yes") & (triage.epss > 0.9)]
print(f"Immediate action: {len(critical)} CVEs")

Defensive Gap Analysis (ATT&CK + Sigma + D3FEND + CAR)

# Find ATT&CK techniques heavily used by APT groups but poorly covered by detections
uses = df[(df.predicate == "uses") & df.subject.str.startswith("G")]
group_usage = uses.groupby("object").subject.nunique().rename("groups_using")

# Count detection sources per technique (Sigma + CAR + D3FEND + ENGAGE)
sigma = df[df.predicate == "detects-technique"].groupby("object").subject.nunique().rename("detections")
d3fend = df[df.predicate == "restricts"].groupby("object").subject.nunique().rename("defenses")

coverage = pd.DataFrame(group_usage).join(sigma).join(d3fend).fillna(0)
gaps = coverage[(coverage.groups_using > 10) & (coverage.detections < 5)]
print(f"High-usage, low-detection techniques: {len(gaps)}")

Supply Chain Risk (GHSA + CVE + EPSS + KEV + ExploitDB)

# Score open-source packages by aggregating risk from linked CVEs
ghsa_cve = df[df.predicate == "related-cve"][["subject", "object"]].rename(columns={"subject": "ghsa", "object": "cve"})
packages = df[df.predicate == "affects-package"][["subject", "object"]].rename(columns={"subject": "ghsa", "object": "pkg"})

epss_scores = df[df.predicate == "epss-score"][["subject", "object"]].copy()
epss_scores["epss"] = epss_scores.object.astype(float)

kev_cves = set(df[(df.predicate == "rdf:type") & (df.object == "KnownExploitedVulnerability")].subject)
exploit_cves = set(df[df.predicate == "exploits-cve"].object)

# Join package → GHSA → CVE → enrichment
risk = packages.merge(ghsa_cve, on="ghsa").merge(epss_scores[["subject", "epss"]], left_on="cve", right_on="subject")
risk["in_kev"] = risk.cve.isin(kev_cves)
risk["has_exploit"] = risk.cve.isin(exploit_cves)
risk["ecosystem"] = risk.pkg.str.split("/").str[0]

# Top ecosystems by high-risk CVE count
high_risk = risk[(risk.epss > 0.5) | risk.in_kev | risk.has_exploit]
print(high_risk.groupby("ecosystem").cve.nunique().sort_values(ascending=False).head(10))

CAPEC → CWE → CVE (Attack Pattern Chain)

capec = load_dataset("s0u9ata/security-kg", "capec")["train"].to_pandas()
cve = load_dataset("s0u9ata/security-kg", "cve")["train"].to_pandas()

# Find CWEs related to SQL Injection (CAPEC-66)
cwe_ids = capec[(capec.subject == "CAPEC-66") & (capec.predicate == "related-weakness")].object.tolist()

# Find CVEs with those CWEs
for cwe_id in cwe_ids:
    related_cves = cve[(cve.predicate == "related-weakness") & (cve.object == cwe_id)].subject.unique()
    print(f"{cwe_id}: {len(related_cves)} CVEs")

D3FEND (Defensive Taxonomy)

ds = load_dataset("s0u9ata/security-kg", "d3fend")
df = ds["train"].to_pandas()

# All 497 defensive techniques in the D3FEND taxonomy
defenses = df[(df.predicate == "rdf:type") & (df.object == "DefensiveTechnique")]
print(f"Defensive techniques: {len(defenses)}")

# Find children of a category (e.g., all techniques under Network Traffic Analysis)
children = df[(df.predicate == "child-of") & (df.object == "NetworkTrafficAnalysis")].subject.tolist()

# Get their names
names = df[df.predicate == "name"][["subject", "object"]]
print(names[names.subject.isin(children)].to_string(index=False))

Source Licensing & Attribution

This dataset is published under the Apache 2.0 license. The underlying source data is provided under various licenses as detailed below. By using this dataset, you agree to comply with each source's respective terms.

Source	License	Attribution
ATT&CK	Custom royalty-free (MITRE)	© The MITRE Corporation. Reproduced and distributed with the permission of The MITRE Corporation.
CAPEC	Custom royalty-free (MITRE)	© The MITRE Corporation. Reproduced and distributed with the permission of The MITRE Corporation.
CWE	Custom royalty-free (MITRE)	© The MITRE Corporation. Reproduced and distributed with the permission of The MITRE Corporation.
CVE	Custom permissive (MITRE)	© The MITRE Corporation. CVE® is a registered trademark of The MITRE Corporation.
CPE / NVD	Public domain (NIST)	This product uses data from the NVD API but is not endorsed or certified by the NVD.
D3FEND	MIT License	© The MITRE Corporation. MITRE D3FEND™ is a trademark of The MITRE Corporation.
ATLAS	Apache 2.0	© MITRE.
CAR	Apache 2.0	© The MITRE Corporation.
ENGAGE	Apache 2.0 (GitHub repo) / Custom restrictive (website ToU)	© The MITRE Corporation. Reproduced and distributed with the permission of The MITRE Corporation. Note: the GitHub repo is licensed Apache 2.0, but the website terms restrict use to internal/non-commercial purposes. Clarification pending with MITRE.
F3	Apache 2.0	© MITRE Engenuity, Center for Threat-Informed Defense.
EPSS	Custom permissive (FIRST)	Jacobs, Romanosky, Edwards, Roytman, Adjerid (2021), Exploit Prediction Scoring System, Digital Threats Research and Practice, 2(3). See first.org/epss.
KEV	Public domain (U.S. Gov)	Source: CISA Known Exploited Vulnerabilities Catalog.
Vulnrichment	CC0 1.0 Universal	Source: CISA Vulnrichment.
GHSA	CC BY 4.0	Source: GitHub Advisory Database. Licensed under CC BY 4.0.
Sigma	Detection Rule License 1.1	Source: SigmaHQ. Licensed under DRL 1.1. Rule author attribution is preserved in triples.
ExploitDB	GPLv2+	Source: OffSec ExploitDB. Derived factual metadata (IDs, CVE mappings, dates) extracted under GPLv2+.
MISP Galaxies	CC0 1.0 / BSD 2-Clause	Source: MISP Project. Dual-licensed under CC0 1.0 and BSD 2-Clause.

License

Apache 2.0 — see Source Licensing & Attribution for individual source terms.