This thesis presents a case study on the application of diffusion models for the synthetic generation of tabular data. Generative models have shown remarkable results in producing synthetic data, and diffusion models are currently one of the most popular generative models . After a brief introduction to the main generative models including Generative adversarial networks, Variational autoencoders and Diffusion models, the thesis shows a case study of a comparison of two diffusion models, TabDDPM and FinDiff. These are applied to two different datasets, and the results are analysed using various metrics, considering statistical properties and privacy aspects. The study demonstrates very high quality in synthetic generation and robust results for privacy protection for mixed tabular data.