With the prevalence of privacy incidents and recurrent leaks, privacy protection has become a concern of users and data protection entities. Previous research in privacy risk prevention has mostly followed the "collect and prevent" philosophy, e.g., by applying anonymization techniques to the user data on the back-end. These approaches neglect the users' right to informational self-determination since the privacy risk analysis and prevention takes place once the data is out of control of the user. In this paper, we present an architecture for user-centered privacy risk detection and quantification framework based on combinatorial, and probabilistic mathematical models coupled with advanced machine learning classifiers. The framework empowers users to take informed privacy prevention actions prior to unwanted revelations.