Makoto Levels Specification

Makoto (誠 — "sincerity") is a specification for describing and incrementally improving data supply chain security, established by industry consensus. It is organized into a series of levels that describe increasing security guarantees.

This is Version 0.1 of the Makoto specification. It defines the provenance levels (L1, L2, L3), as well as recommended attestation formats for producing DBOMs (Data Bills of Materials).

Overview

Makoto is a security framework that establishes trust in data artifacts through verifiable attestations. It brings SLSA-style assurance levels to data pipelines, addressing the unique challenges of data supply chains while sharing SLSA's philosophy of progressive security levels.

Core Principle: Every data artifact should be accompanied by attestations that document its origin, transformations, and lineage—enabling consumers to verify the data's integrity and provenance. The resulting DBOM proves where your data came from.

Why Makoto?

While SLSA effectively secures software build pipelines, data pipelines have fundamentally different characteristics:

Characteristic	Software (SLSA)	Data (Makoto)
Artifact Type	Discrete builds	Streams and batches
Transformations	Single build step	Multi-stage ETL pipelines
Granularity	File/package	Record/window/table
Privacy	Generally public	May contain PII
Velocity	Per-release	Millions of events/sec
Output	SBOM	DBOM

Makoto Levels

Makoto defines three progressive levels, each building on the previous with stronger guarantees.

L1 Provenance Exists

"Documentation exists for data origin and processing."

Requirement	Description
Attestation Exists	Machine-readable attestation accompanies the data artifact
Origin Documented	Source of data is recorded (may be self-attested)
Transforms Documented	Processing steps are listed (best-effort completeness)
Format Compliance	Uses Makoto attestation schema with valid predicate type

Adoption Difficulty: Low — Can be implemented with manual documentation or simple scripting.

L2 Provenance is Authentic

"Attestations are cryptographically signed and tamper-evident."

Requirement	Description
All L1 Requirements	Plus additional controls below
Signed Attestations	Digital signatures using ECDSA P-256 or equivalent
Tamper-Evident	Consumers can verify signature validity against known keys
Timestamp Binding	Attestations include verifiable timestamps (RFC 3161 or similar)
Hash Chaining	Each transform attestation references cryptographic hash of inputs

Adoption Difficulty: Medium — Requires key management infrastructure. Can use Sigstore.

L3 Provenance is Unforgeable

"Attestations are generated by isolated infrastructure that processing code cannot influence."

Requirement	Description
All L2 Requirements	Plus additional controls below
Isolated Signing	Signing keys stored in secure enclave, inaccessible to processing code
Platform-Generated	Attestations created by trusted control plane, not user code
Deterministic Hashing	Data hashes computed by platform, not influenced by tenant
Immutable Audit Trail	All attestation operations logged to append-only store

Adoption Difficulty: High — Requires platform support with TEE or HSM.

Attestation Model

Makoto attestations use the in-toto attestation framework with Makoto-specific predicate types.

Attestation Structure

{
  "_type": "https://in-toto.io/Statement/v1",
  "subject": [{
    "name": "dataset:customer_transactions",
    "digest": {
      "sha256": "abc123...",
      "recordCount": "1000000"
    }
  }],
  "predicateType": "https://makoto.dev/origin/v1",
  "predicate": {
    // Predicate-specific content
  }
}

Predicate Types

Predicate Type	Purpose	Used For
`makoto.dev/origin/v1`	Data origin and collection	Initial data ingestion
`makoto.dev/transform/v1`	Data transformation	ETL, filtering, aggregation
`makoto.dev/stream-window/v1`	Stream window attestation	High-throughput streaming

Data Bill of Materials (DBOM)

A DBOM documents the complete lineage of a data artifact, similar to an SBOM for software.

{
  "dbomVersion": "1.0.0",
  "dataset": {
    "name": "ml_training_dataset_v3",
    "version": "3.0.0",
    "makotoLevel": "L2"
  },
  "sources": [
    {
      "name": "customer_transactions",
      "attestationRef": "https://attestations.example.com/origin/abc",
      "makotoLevel": "L2"
    }
  ],
  "transformations": [
    {
      "order": 1,
      "name": "Anonymize PII",
      "attestationRef": "https://attestations.example.com/transform/def"
    }
  ]
}

Verification

Consumers verify DBOMs and Makoto attestations through the following steps:

1. Fetch Attestation

→

2. Verify Signature

→

3. Check Data Hash

→

4. Verify Lineage

Fetch attestation — Retrieve attestation from registry or alongside data
Verify signature — Validate cryptographic signature against trusted keys
Check data hash — Compute hash of data, compare to attestation subject
Verify lineage — Recursively verify input attestations