Linux Security Meanfield: Security-Relevant Commits from 22 Linux Repositories

A commit-keyed corpus of security-relevant changes across 22 Linux base-system repositories. The dataset contains 2,254 rows of CVE-dossiered fixes and 21,609 rows of non-CVE security signal commits, unified on a single schema. It was created by author mjbommar as the Phase-1 release artifact for a mean-field survey paper.

Use Cases

• Analyzing patterns in CVE-fixing commits based on the CVE-dossiered corpus.
• Studying security-relevant code changes that lack CVE assignments based on the non-CVE signal commits.
• Training models for automated security patch identification based on the unified commit schema.
• Conducting longitudinal studies of Linux kernel security evolution based on the 22-repository scope.

Strengths

• Contains 23,863 total rows of security-relevant commits.
• Scope includes 22 distinct Linux base-system repositories.
• Explicitly separates CVE-dossiered fixes (2,254 rows) from broader security signals (21,609 rows).

Limitations

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

Provenance

Source

22 Linux base-system repositories.

Collection Method

Commit-keyed corpus extraction, likely from version control history.

Time Range

null

Freshness

Last updated 2026-04-22 12:35:55; freshness should be verified.

Geography

null