CVE Security Patch Backport Code Generation Dataset

Per-hunk code generation dataset derived from openSUSE Build Service maintenance patches. It provides vulnerable code regions and descriptions of upstream CVE fixes for training models to output fixed code versions. The dataset was created by openSUSE and last updated in April 2026.

Use Cases

• Generate fixed code versions from vulnerable source code regions and CVE fix descriptions.
• Train models for automated CVE patch backporting using per-hunk code completion.
• Evaluate LLM performance on security-specific code generation tasks using programmatic diff outputs.

Strengths

• Derived from real-world openSUSE Build Service maintenance patches.
• Focuses on per-hunk generation to avoid format-sensitivity issues.

Limitations

• Unknown row count and dataset size.
• Unknown column structure and specific data features.
• Potential bias towards openSUSE-specific software packages.

Provenance

Source

openSUSE Build Service maintenance patches.

Collection Method

Derived from open source maintenance patches.

Freshness

Last updated April 2026.