Skip to content

Evidence Management Standard

ID STD-11 · Status: Template. Adapt it and have it approved in your organization.
Owner AppSec lead. Platform team operates.
Related Evidence model, ADR-004, ADR-016, ADR-017, Data Retention Policy

1. Purpose

Defines what counts as security evidence, how it is produced, verified, stored, aged and retained, so that every important security claim can be backed by evidence.

2. Requirements

ID Requirement
EVM-01 Every control in the catalogue shall define its evidence types, minimum confidence and proof statement. The catalogue validator enforces it.
EVM-02 Evidence shall carry: type, application, scope, subject (commit/digest/setting), producer, tool and version, policy, observation time, result, confidence, raw-artefact hashes, and the run reference where applicable
EVM-03 Declared information (C0) never satisfies a control. Configuration presence (C1) never satisfies an activity control.
EVM-04 Pipeline evidence is accepted only from trusted producers (config/trusted-producers.yaml) after verification of workflow identity, attestation, digests and commit binding
EVM-05 Manual evidence records the reviewer from their authenticated session. M2 requires a reviewer outside the owning team. Attachments are malware-scanned and hashed.
EVM-06 Evidence records and raw artefacts are immutable. Corrections are made by revocation or supersession records.
EVM-07 Evidence freshness is computed by the freshness engine per evidence type (config/evidence-types.yaml). Expired evidence does not satisfy controls.
EVM-08 Evidence integrity (hash chain, blob hashes) is verified daily. Failures are P1 platform incidents.
EVM-09 Evidence is retained per the Data Retention Policy. WORM retention is not locked until that policy is approved.
EVM-10 Evidence for public repositories is internal and never published with the source
EVM-11 Evidence may be revoked by AppSec (for example, a defective tool version or a compromised producer). Revocations are audited and trigger re-evaluation.
EVM-12 Access to raw artefacts is role-scoped, via short-lived single-blob links, and audited

3. Confidence levels

C0 declared · C1 configured · C2 executed · C3 attested/authoritative · M1 manual-self · M2 manual-independent (definitions).

4. Audit use

Auditors receive time-bound auditor role access and can traverse: control state → evaluation → evidence → raw artefact → hash → daily checkpoint (also present in Sentinel).