Faceted Data
The article by Read Schmitz et al (2016) describes faceted values as values that present different behavior according to the privilege of the observers and faceted data as a way of organizing data so that it can be viewed and manipulated in different ways. Faceted evaluation is a technique for simulating secure multi-execution with a single process, using special faceted values that contain both a public view and a private view of the data, meaning a single execution can provide many of the same guarantees that secure multi-execution provides while achieving better performance.
I think this is a good approach to protect system from data leakage as you are sure that only authorized users with recommended privileges can view what is recommended to them. According to Schmitz et al (2016), faceted values can be incorporated within existing language via library support and this means faceted data is flexible and can adapt to the needs of an application.
However, prior work to implement faceted values is a complex processes and requires development of new language semantics and also the implementation can be ineffective if poorly designed.
A basic outline design of creating this in python is shown below;
