The frontiers between humans and the digital world are tightening. An unprecedented and rapidly increasing amount of individual data ends up nowadays into well-structured and consistent databases. Privacy-preserving data publishing models and algorithms are an attempt for benefiting collectively from this wealth of data while still preserving the privacy of individuals. However, few works have focused on their implementation issues, leading to highly vulnerable practical scenarios. This thesis tackles precisely this issue, benefiting from the rise of portable, large capacity, and tamper-resistant devices used as personal data servers. Its contribution is a set of correct and secure protocols permitting to perform traditional non-interactive privacy- preserving data publishing algorithms and models in an original context where each individual manages her data autonomously through her own secure personal data server.