> ## Documentation Index
> Fetch the complete documentation index at: https://docs.zdlake.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Security data model

> How vulnerabilities, advisories, software, weaknesses, attack patterns, assessments, and source evidence relate.

Zero separates concrete security issues, publications, affected software, reusable taxonomies, and source evidence. This keeps related concepts connected without treating them as interchangeable.

## Layered ontology

```mermaid theme={null}
flowchart LR
  subgraph security_records["Security records"]
    advisory["Advisory"]
    vulnerability["Vulnerability"]
    software["Software"]
  end
  subgraph reusable_taxonomies["Reusable taxonomies"]
    weakness["Weakness"]
    attack_pattern["Attack pattern"]
  end
  subgraph identifiers_and_evidence["Identifiers and evidence"]
    cve["CVE"]
    assessment["Assessment"]
    exploitation_evidence["Exploitation evidence"]
    remediation["Remediation"]
  end
  advisory <--> vulnerability
  vulnerability --> software
  weakness --> vulnerability
  attack_pattern --> weakness
  vulnerability --> cve
  assessment --> vulnerability
  exploitation_evidence --> vulnerability
  remediation --> vulnerability
  remediation --> software
```

The diagram is a visual summary. The tables below define every concept and relationship in text.

## Graph relationships

| From | Relationship | To |
| - | - | - |
| Advisory | Bidirectional relationship | Vulnerability |
| Vulnerability | Directed relationship | Software |
| Weakness | Directed relationship | Vulnerability |
| Attack pattern | Directed relationship | Weakness |
| Vulnerability | Directed relationship | CVE |
| Assessment | Directed relationship | Vulnerability |
| Exploitation evidence | Directed relationship | Vulnerability |
| Remediation | Directed relationship | Vulnerability |
| Remediation | Directed relationship | Software |

## Core concepts

| Concept | Definition |
| - | - |
| [Vulnerability](https://docs.zdlake.com/concepts/vulnerabilities.md) | A concrete security flaw that can be exploited. A CVE is one possible public identifier for a vulnerability. |
| [Advisory](https://docs.zdlake.com/concepts/advisories.md) | A publisher-issued security notice that may cover multiple vulnerabilities, affected software, and remediations. |
| [Weakness](https://docs.zdlake.com/concepts/weaknesses.md) | A recurring class of design or implementation error that can cause vulnerabilities. |
| [Software](https://docs.zdlake.com/concepts/software.md) | A package, application, operating system, appliance, hardware device, or firmware that may be affected by a vulnerability. |
| [Attack pattern](https://docs.zdlake.com/concepts/attack-patterns.md) | A reusable description of how adversaries may exploit weaknesses. It is not a vulnerability. |

## Important cardinalities

| Relationship | Cardinality | Meaning |
| - | - | - |
| Vulnerability and CVE ID | A vulnerability has zero or one CVE ID | A vulnerability may have zero or one CVE ID. |
| Advisory and Vulnerability | An advisory covers zero or more vulnerabilities | An advisory may cover zero or more vulnerabilities. |
| Vulnerability and Advisory | A vulnerability appears in zero or more advisories | A vulnerability may appear in zero or more advisories. |
| Vulnerability and Software | A vulnerability affects zero or more software identities | A vulnerability may affect zero or more software identities. |
| Weakness and Vulnerability | Many-to-many | A weakness may classify zero or more vulnerabilities, and a vulnerability may map to zero or more weaknesses. |
| Attack pattern and Weakness | Many-to-many | An attack pattern may exploit zero or more weaknesses, and a weakness may relate to zero or more attack patterns. |
| Assessments and exploitation evidence to Vulnerability | Zero or more per vulnerability | Exploitation evidence and assessments may attach to a vulnerability without changing its publication state. |

These cardinalities describe customer-visible meaning, not database implementation.

## Identifiers are not entities

| Identifier | What it identifies |
| - | - |
| [CVE ID](https://docs.zdlake.com/concepts/security-data-model.md#identifiers-are-not-entities) | A publicly disclosed vulnerability record. |
| [Advisory ID](https://docs.zdlake.com/concepts/advisories.md) | A publisher's versioned security notice. |
| [CWE ID](https://docs.zdlake.com/concepts/weaknesses.md) | A weakness class. |
| [CAPEC ID](https://docs.zdlake.com/concepts/attack-patterns.md) | An attack pattern. |
| [PURL](https://docs.zdlake.com/concepts/software.md#identity-rules) | Package-shaped software. |
| [CPE](https://docs.zdlake.com/concepts/software.md#identity-rules) | A product, application, operating system, hardware device, or firmware. |
| [Source-scoped identifier](https://docs.zdlake.com/concepts/software.md#identity-rules) | A publisher-specific software identity when no standard coordinate is available. |

## Assessments and exploitation evidence

An assessment is an issuer-owned measurement or decision about a vulnerability. CVSS, EPSS, SSVC, and CISA KEV inclusion retain their method, value, issuer, and assessment time rather than becoming one blended risk fact.

Exploitation evidence is a source-backed claim about exploit availability or observed exploitation. Public proof-of-concept availability and active exploitation are separate states, and missing evidence is not evidence that exploitation has not occurred.

Affected assertions relate vulnerabilities to software with an affected, fixed, unaffected, or investigating version scope. Remediation remains scoped to the relevant vulnerability and software rather than applying to every related release.

## Software and releases

Software is a versionless identity for a package, application, operating system, appliance, hardware device, or firmware. A Software record can have many registry-observed release children. A release records an exact version and registry metadata; it does not create another Software identity.

Release assessments join a registry-observed release to a normalized affected assertion. The API aggregates those rows by vulnerability or advisory using explicit conflict semantics: agreeing sources produce one status and source count, while disagreement produces `unknown` with `conflicting_sources`. Release and nested subject lists are deterministically capped at 100 and expose totals and truncation metadata. The registry supplies release observations, while vulnerability databases and publisher advisories supply the affected-version claims being evaluated; general web search does not create release records.

## Derived classifications

Zero-day, actively exploited, newly disclosed, and high priority are computed views over entities, evidence, assessments, and time. They are not entity types.

A vulnerability is classified as exploited as a zero-day only when accepted source evidence reports exploitation in the wild before an official vendor fix was available. A CVE assigned later does not erase that history.

## Provenance

Zero preserves the derivation chain from a fetched artifact to an immutable record version, its source claims, and the resolved values returned to customers. This follows the W3C PROV model while keeping the product language focused on sources and evidence.

> [card] [Provenance](https://docs.zdlake.com/concepts/provenance.md)
>
> Trace normalized values back to immutable source evidence.

> [card] [Data sources](https://docs.zdlake.com/concepts/data-sources.md)
>
> See the public datasets that contribute records and evidence.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.