Generate and certify synthetic datasets

Five input modes. Five generation engines. Five certificate levels. Every output is SHA-256 fingerprinted and Ed25519 signed — independently verifiable without contacting CertifiedData.

Generate now →Browse registry Verify a certificate

Live execution — no account required

Live execution — sandboxPOST /api/synth/public/generate

Template

Engine

Rows (max 1,000 sandbox)

10 anonymous runs / 24h per IP · sandbox.cert.v1 issued · 7-day TTL

Sandbox certificates carry a scert_ prefix. Claim and upgrade to a persistent cert.v1 once you have a free account. How certification works →

Input modes

Template

nonePOST /api/synth/public/generate

Select a domain template (fraud, healthcare, finance, HR…). CertifiedData applies a pre-validated schema. Fastest path to a certified sandbox dataset. Only engine available without account: light.

{"templateId":"fraud_detection","rows":500,"engine":"light"}

Prompt → Schema Plan

Bearer requiredPOST /api/genesis/plan

Describe the dataset in natural language. The API returns a typed JSON schema plan you review before executing. Pipeline: prompt → AI schema planner → JSON schema → generate. Prompt text never drives execution directly.

{"prompt":"500 rows of synthetic patient encounters with diagnosis codes and insurance claims"}

Schema-run

Bearer requiredPOST /api/genesis/schema-run

Execute generation from a saved schema ID. First create a schema via POST /api/genesis/schema, then trigger runs against it. Returns a certified output dataset.

{"schema_id":"schema_xxxx","rows":1000,"engine":"ctgan"}

Manifest upload

Bearer requiredPOST /api/manifest/upload

Upload a manifest JSON describing the dataset specification, provenance metadata, and generation intent. Used for reproducible multi-job pipelines and CI-driven certification.

{"version":"1","domain":"finance","columns":[{"name":"transaction_id","type":"uuid"},{"name":"amount","type":"float","min":0.01,"max":50000}]}

Notary / Certify

3 free (anonymous) or BearerPOST /api/notary/certify

Submit any existing artifact (file or hash) for notarization. Issues an Ed25519 attestation certificate without generating new data. First 3 certifications free per account.

{"artifact_type":"dataset","metadata":{"source":"ctgan-v0.7","rows":10000}}

Generation engines

Engine	Plan	Description
light	Free (anonymous)	Fast statistical sampling. No training. Only engine available without an account.
gaussian	Build+	Gaussian copula. Column-wise distributions + correlation matrix. Requires account.
ctgan	Trust+	Conditional Tabular GAN. High-fidelity structural replication. No privacy guarantee.
hybrid	Govern+	Gaussian with learned cross-column correlations. Better marginal fidelity.
dp-ctgan	Govern+	CTGAN with active differential privacy accounting and noise injection. Epsilon-bounded.

DP-CTGAN uses active differential privacy accounting with epsilon-bounded noise injection. Do not use light/gaussian/hybrid engines and claim differential privacy — no noise is added.

Certificate levels

Level	Plan	TTL	Includes
sandbox.cert.v1	Anonymous	7 days	scert_ prefix · not in production registry · claim-and-upgrade available
cert.v1	Free+	Permanent	Registry-listed · SHA-256 + Ed25519 · public verify URL
cert.v2	Build+	Permanent	cert.v1 + policy metadata, generation params
cert.v3	Trust+	Permanent	cert.v2 + audit vault, dual-party signing
cert.v4	Govern+	Permanent	cert.v3 + full chain-of-custody, EU AI Act Article 10/12 fields

Capability matrix

Action	Anonymous	Authenticated	Production	Public Proof
Template generation	✓ 10/24h · 1k rows	✓ 5 jobs/mo · 10k rows	✓ unlimited	✓ verify URL
Prompt → spec → generate	✗ auth required	✓ Build+	✓ Build+	✓ cert.v1+
Schema-controlled gen	✗ auth required	✓ Build+	✓ Build+	✓ cert.v2+
Manifest upload	✗ auth required	✓ Build+	✓ Build+	✓ cert.v1+
Upload + synthesize	✗ auth required	✓	✓	✓ cert.v1+
Notary / attest	✓ 3 free	✓ plan quota	✓ unlimited	✓ verify URL
CTGAN engine	✗ plan required	✗ Trust+ required	✓ Trust+	✓ cert.v1+
DP-CTGAN engine	✗ plan required	✗ Govern+ required	✓ Govern+	✓ cert.v4
cert.v1 issuance	✗ sandbox only	✓ free+	✓	✓ registry
cert.v4 issuance	✗	✗ Govern+ required	✓ Govern+	✓ full chain

API examples

Anonymous sandbox generation

curl -X POST https://certifieddata.io/api/synth/public/generate \
  -H "Content-Type: application/json" \
  -d '{
    "templateId": "fraud_detection",
    "rows": 100,
    "engine": "light"
  }'

Schema-controlled generation (authenticated)

curl -X POST https://certifieddata.io/api/generate/schema \
  -H "Authorization: Bearer $CERTIFIEDDATA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "schema": {
      "type": "object",
      "properties": {
        "transaction_id": {"type": "string", "format": "uuid"},
        "amount": {"type": "number", "minimum": 1, "maximum": 50000},
        "is_fraud": {"type": "boolean"}
      }
    },
    "rows": 1000,
    "engine": "ctgan",
    "cert_level": "cert.v2"
  }'

Claim and upgrade a sandbox cert

curl -X POST https://certifieddata.io/api/sandbox/claim-and-upgrade \
  -H "Authorization: Bearer $CERTIFIEDDATA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"sandbox_cert_id":"scert_xxxx"}'

Related surfaces

Generate Certification guide Verify a certificate Registry Dataset marketplace Workflows Safety & boundaries