AIBOM Schema Achieves 98.7% Reproducibility for AI Supply Chains
The CycloneDX standard has been enhanced with a new Artificial Intelligence Bill of Materials (AIBOM) schema, designed to document AI-specific provenance, model lineage, and disclosure metadata. This framework employs structured schema engineering, cryptographic validation, and agent-driven automation to ensure verifiable software provenance. An autonomous AI pipeline conducts ongoing environment inspections, enriches vulnerability data, and audits for reproducibility. Results from empirical evaluations indicate a reproducibility fidelity of 98.7%, a vulnerability match precision of 96.2%, and a significant 63% decrease in manual oversight within containerized analytic workflows.
Key facts
- AIBOM schema extends CycloneDX standard for AI provenance
- Framework uses cryptographic validation and agent-driven automation
- Autonomous pipeline performs environment inspection and vulnerability enrichment
- 98.7% reproducibility fidelity achieved
- 96.2% vulnerability match precision
- 63% reduction in manual oversight
- Published on arXiv as 2605.19755v1
- Addresses challenges in AI software supply chain transparency
Entities
Institutions
- arXiv