This repository contains the code and data for the paper “Quantifying Privacy Risks in Synthetic Data: A Study on Black-Box Membership Inference”. It enables full reproducibility of the empirical study that evaluates privacy–utility trade-offs in synthetic tabular data under multiple black-box membership inference attacks (MIAs). Using the Adult, COMPAS, and Southern German Credit datasets, the code repeatedly generates synthetic datasets (30 train/test splits) with four generators (Mixup, TVAE, CTGAN, CTAB-GAN+), measures downstream utility via XGBoost performance, and quantifies privacy risk using several MIAs (distance-, distribution-, Monte Carlo-based, and DOMIAS) together with common privacy metrics (DCR, NNDR, and Privacy Loss).