Cybersecurity Dataset Fenrir V2.1: 99,870 Defensive Instruction-Tuning Triples

Alican Kiraz created a dataset of 99,870 high-quality system, user, and assistant triples for supervised fine-tuning in defensive cybersecurity. The dataset is licensed under Apache-2.0 and is described as production-ready. Its scope includes OWASP Top 10, MITRE ATT&CK, NIST CSF, CIS Controls, ASD Essential 8, modern authentication, SSL/TLS, Cloud & DevSecOps, Cryptography, and AI Security.

Use Cases

• Train language models for defensive cybersecurity response based on the described OWASP Top 10 and MITRE ATT&CK frameworks.
• Fine-tune assistants for secure coding guidance based on the described CIS Controls and ASD Essential 8 scope.
• Develop instructional agents for cloud security and DevSecOps practices based on the described scope.
• Create educational tools for modern authentication protocols like OAuth 2, OIDC, and SAML based on the described dataset content.

Strengths

• Contains 99,870 high-quality instruction-tuning triples.
• Explicitly licensed under Apache-2.0 for production use.
• Scope is defined across multiple major cybersecurity frameworks and topics.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but other scale metrics like file size and format are unknown.
• Freshness should be verified as the last update timestamp is 2026-05-30 07:35:15.

Provenance

Source

Author Wuluyu on Hugging Face.

Collection Method

Likely created for instruction-tuning in cybersecurity; specific gathering method is not detailed.

Freshness

Last updated 2026-05-30 07:35:15.