-
-
Notifications
You must be signed in to change notification settings - Fork 78
Description
Feature Request: Perspectives: Domain-Specific Views into CycloneDX Data
Summary
Introduce a perspectives construct that enables domain-specific audiences to define curated views into CycloneDX BOMs or API responses. Each perspective identifies relevant data types using JSONPath expressions and provides domain-specific terminology mappings, allowing diverse stakeholders to interpret CycloneDX data through their own conceptual lens.
Background and Motivation
This proposal originates from discussions within the Threat Modeling Working Group, where the concept of "perspectives" emerged as a means to provide different views into a blueprint or model. Different audiences have distinct concerns and use different terminology to describe the same underlying concepts. For example:
| Audience | Sample Concerns |
|---|---|
| Application Security (AppSec) | Vulnerabilities, dependencies, code provenance, threat models |
| Operational Security (OpSec) | Runtime configurations, deployment topology, secrets management |
| Physical Security | Hardware components, device locations, tamper evidence |
| Cyber Security | Cryptographic assets, attack surfaces, threat models |
| AI/ML Engineers | Model lineage, training data, foundation models, bias considerations |
| Legal/Compliance | Patents, licensing, export controls, intellectual property |
Furthermore, different domains use different terminology for equivalent CycloneDX concepts:
| CycloneDX Concept | AI/ML Terminology | Security Terminology |
|---|---|---|
pedigree.ancestors |
Foundation model, Base model | Upstream source |
modelCard |
Model transparency report | — |
formulation |
Training pipeline | Build provenance |
The perspectives feature proposes to address these challenges by:
- Enabling domain experts to define which parts of a CycloneDX data are relevant to their discipline
- Providing a mechanism to map CycloneDX terminology to domain-specific nomenclature
- Supporting tooling that can present filtered, contextualised views of BOM data