Cardiovascular diseases are leading cause of death worldwide. Risk assessment is extremely vital in prevention and treatment, and should be done early and accurately. Machine learning is commonly applied in the past few years to forecast cardiovascular risk based on healthcare data. This information comprises of electronic health records, medical images, wearable devices and genetic information. At the same time, healthcare systems are moving toward distributed environments that use mobile devices, cloud platforms, and the Internet of Medical Things. These systems improve access to care but also create serious problems related to data security and patient privacy. This review summarizes machine learning methods used for cardiovascular risk assessment in distributed healthcare systems. It also explains the main security and privacy challenges in these environments. In addition, the review discusses secure machine learning approaches such as federated learning and differential privacy. Finally, it highlights key research gaps and future directions to support safe and reliable use of machine learning in cardiovascular healthcare.