Cybersec Distillation Data for Fine-Tuning LLM Agents with Tool-Calling

14,491 CVEs enriched through CWE to CAPEC to ATT&CK to D3FEND chains form part of this training data. The dataset was distilled from DeepSeek-R1 using a multi-stage pipeline and includes 925 HackTricks pentesting entries and 280 Kali tool descriptions. Author neilopet last updated it on Hugging Face in February 2026.

Use Cases

• Fine-tune cybersecurity LLM agents for vulnerability analysis based on enriched CVE data.
• Train models on penetration testing methodologies based on the 925 HackTricks entries.
• Develop tool-calling capabilities for security tools based on the 280 Kali tool descriptions.

Strengths

• Includes 14,491 CVEs enriched through a multi-stage mapping chain (CWE -> CAPEC -> ATT&CK -> D3FEND).
• Contains 925 entries from HackTricks covering pentesting methodologies.
• Provides 280 tool descriptions from Kali Linux with usage guidance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment.

Provenance

Source

Distilled from DeepSeek-R1; sources include NVD CVEs, HackTricks, and Kali tools.

Collection Method

Multi-stage pipeline for collecting, enriching, and generating cybersecurity Q&A data.

Time Range

null

Freshness

Last updated 2026-02-15 20:43:05; freshness should be verified.

Geography

null