In the previous two chapters we discussed centralized differential privacy (DP), where there is a trusted data curator who collects data from individuals and applies different techniques to obtain differentially private statistics about the population. Then the curator publishes privacy-preserving statistics about this population. However, these techniques are unsuitable when individuals do not completely trust the data curator. Hence, various techniques to satisfy DP in the local setting have been studied to eliminate the need for a trusted data curator. In this chapter we will walk through the concept, mechanisms, and applications of the local version of DP, local differential privacy (LDP).
This chapter will mainly look at how LDP can be implemented in ML algorithms by looking at different examples and implementation code. In the next chapter we’ll also walk you through a case study of applying LDP naive Bayes classification for real-world datasets.
