This chapter investigates ways to deal with privacy rules when modeling preferences of users in recommender systems based on collaborative filtering. It argues that it is possible to find a good compromise between quality of predictions and protection of personal data. Thus, it proposes a methodology that fulfills with strictest privacy laws for both centralized and distributed architectures. The authors hope that their attempts to provide an unified vision of privacy rules through the related works and a generic privacy-enhancing procedure will help researchers and practitioners to better take into account the ethical and juridical constraints as regards privacy protection when designing information systems.