BigGAN is a powerful generative adversarial network capable of producing high-fidelity synthetic images. However, deploying GANs in distributed environments involving sensitive data poses significant privacy risks due to potential data memorization. To address this challenge, Differentially Private BigGAN has been integrated within a Hybrid Federated–Centralized Learning (HFCL) framework. Differential privacy is enforced through gradient clipping and Gaussian noise addition, while BigGAN stability is enhanced using spectral normalization, orthogonal regularization, self-attention, and gradient penalty. In the proposed HFCL setup, local client training is combined with centralized aggregation via a secure socket-based communication protocol. The framework is evaluated using Fréchet Inception Distance (FID), Kernel Inception Distance (KID), Perceptual Path Length (PPL), Peak Signal-to-Noise Ratio (PSNR), and Structural Similarity Index Measure (SSIM) metrics. Experimental results demonstrate that the proposed DP-BigGAN-HFCL framework achieves high-fidelity image generation while significantly reducing data leakage risks, making it suitable for privacy-sensitive applications such as biometrics and healthcare.