Fair and Privacy-Preserving Synthetic Data Generation via Clustering-Based Variational Autoencoder and Adversarially Debiased Wasserstein Generative Adversarial Networks with Gradient Penalty