Synthesize

Generate a privacy-safe synthetic patient cohort

Data Source

Rows to Generate: 200

Epsilon (ε): 1.0

DP Mechanism

Seed (optional)

Synthesis Pipeline

①

Gaussian Copula fit

SDV learns joint column distributions from real data

②

L2 row clipping (FR-009-A)

Clip rows to ‖x‖₂ ≤ √p before covariance estimation

③

DP noise injection (FR-010)

Gaussian (ε,δ)-DP or Wishart pure ε-DP added to Σ

④

Temporal constraints (FR-019-A)

Swap col_end/col_start pairs where end < start

⑤

DCR/NNDR purge

Exact clones removed; Gower distance checked vs real

⑥

EDPB three-criteria test

Singling-out, linkage, inference adversarial checks

Compliance & Guarantees

EHDS 2026European Health Data Space compliant synthetic data standard

HIPAA §164.514(b)Expert Determination method — re-ID risk ≤ 0.09 (9%) threshold

(ε,δ)-DPMathematically proven differential privacy with composable budget

EDPB Op. 28/2024Three-criteria anonymisation test per para 40 (singling-out, linkage, inference)